[clang] [HLSL] Disable implicit constructors for user-defined structs/classes (PR #194989)
Helena Kotas via cfe-commits
cfe-commits at lists.llvm.org
Sat Jun 13 10:42:43 PDT 2026
https://github.com/hekota updated https://github.com/llvm/llvm-project/pull/194989
>From 9933d7acf5e34762631eb73d3de902fb17b45e0f Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Wed, 29 Apr 2026 15:49:22 -0700
Subject: [PATCH 01/22] [HLSL] Disable implicit constructors for user-defined
structs/classes
---
clang/include/clang/AST/DeclCXX.h | 66 ++++++--
clang/lib/AST/Expr.cpp | 3 +-
clang/lib/CodeGen/CGDecl.cpp | 8 +-
clang/lib/CodeGen/CGHLSLRuntime.cpp | 13 +-
clang/lib/Sema/SemaExpr.cpp | 28 +++-
clang/lib/Sema/SemaHLSL.cpp | 21 ++-
clang/lib/Sema/SemaInit.cpp | 12 +-
clang/lib/Sema/SemaOverload.cpp | 14 +-
clang/test/AST/HLSL/cbuffer.hlsl | 2 +-
clang/test/AST/HLSL/matrix-constructors.hlsl | 2 +
.../HLSL/semantic-output-struct-shadow.hlsl | 4 +-
.../test/AST/HLSL/semantic-output-struct.hlsl | 4 +-
.../BasicFeatures/ArrayElementwiseCast.hlsl | 4 +-
.../CodeGenHLSL/BasicFeatures/InitLists.hlsl | 156 +++++++++++++-----
.../BasicFeatures/MatrixElementTypeCast.hlsl | 6 +-
.../BasicFeatures/StructElementwiseCast.hlsl | 10 +-
.../BasicFeatures/VectorElementwiseCast.hlsl | 10 +-
.../CodeGenHLSL/builtins/hlsl_resource_t.hlsl | 8 +-
.../StructuredBuffers-subscripts.hlsl | 16 +-
.../resources/cbuffer_struct_passing.hlsl | 122 ++++++++++----
.../resources/resources-in-structs.hlsl | 2 +-
.../semantics/semantic-struct-2-output.hlsl | 4 +-
.../CodeGenHLSL/this-assignment-overload.hlsl | 52 +++---
clang/test/CodeGenHLSL/this-assignment.hlsl | 47 ++++--
.../BuiltIns/WaveActiveAllTrue-errors.hlsl | 2 +-
.../BuiltIns/WaveActiveAnyTrue-errors.hlsl | 2 +-
.../BuiltIns/WaveActiveBallot-errors.hlsl | 2 +-
.../BuiltIns/WaveActiveCountBits-errors.hlsl | 2 +-
.../BuiltIns/dot4add_i8packed-errors.hlsl | 2 +-
.../BuiltIns/dot4add_u8packed-errors.hlsl | 2 +-
.../Language/AggregateSplatCast-errors.hlsl | 4 +-
.../Language/ElementwiseCast-errors.hlsl | 7 +-
.../SemaHLSL/Language/ElementwiseCasts.hlsl | 1 +
clang/test/SemaHLSL/Language/InitListAST.hlsl | 6 +
clang/test/SemaHLSL/Language/InitLists.hlsl | 10 +-
clang/test/SemaHLSL/prohibit_pointer.hlsl | 2 +-
36 files changed, 455 insertions(+), 201 deletions(-)
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index 2af396f025c93..fe2bc0dd628c3 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -759,19 +759,40 @@ class CXXRecordDecl : public RecordDecl {
needsImplicitDefaultConstructor();
}
+ // Used by HLSL to determine if a record is a built-in implicit HLSL
+ // struct/class or a user-defined one. User-defined HLSL records cannot
+ // have ctors, dtors, or overloaded operators, while implicit built-in
+ // HLSL records such as resource classes can. It would be nice to use the
+ // isImplicit() methods to determine that, but this flag is not propagated
+ // to template-instanticated classes.
+ //
+ /// Determines whether this class has any user provided special members.
+ bool hasUserProvidedSpecialMembers() const {
+ return data().UserDeclaredSpecialMembers &
+ (SMF_MoveConstructor | SMF_MoveAssignment | SMF_Destructor |
+ SMF_CopyAssignment | SMF_CopyConstructor) ||
+ data().UserDeclaredConstructor ||
+ data().UserProvidedDefaultConstructor;
+ }
+
/// Determine if we need to declare a default constructor for
/// this class.
///
/// This value is used for lazy creation of default constructors.
bool needsImplicitDefaultConstructor() const {
- return (!data().UserDeclaredConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
- // FIXME: Proposed fix to core wording issue: if a class inherits
- // a default constructor and doesn't explicitly declare one, one
- // is declared implicitly.
- (data().HasInheritedDefaultConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor));
+ return ((!data().UserDeclaredConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
+ // FIXME: Proposed fix to core wording issue: if a class inherits
+ // a default constructor and doesn't explicitly declare one, one
+ // is declared implicitly.
+ (data().HasInheritedDefaultConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor))) &&
+ // In HLSL, only built-in records like resources classes can have
+ // constructors.
+ (!getLangOpts().HLSL ||
+ (isLambda() && lambdaIsDefaultConstructibleAndAssignable()) ||
+ hasUserProvidedSpecialMembers());
}
/// Determine whether this class has any user-declared constructors.
@@ -797,7 +818,11 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// constructor to be lazily declared.
bool needsImplicitCopyConstructor() const {
- return !(data().DeclaredSpecialMembers & SMF_CopyConstructor);
+ return !(data().DeclaredSpecialMembers & SMF_CopyConstructor) &&
+ // In HLSL, only built-in records like resources classes can have
+ // constructors.
+ (!getLangOpts().HLSL || isLambda() ||
+ hasUserProvidedSpecialMembers());
}
/// Determine whether we need to eagerly declare a defaulted copy
@@ -893,8 +918,11 @@ class CXXRecordDecl : public RecordDecl {
return !(data().DeclaredSpecialMembers & SMF_MoveConstructor) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveAssignment() &&
- !hasUserDeclaredDestructor();
+ !hasUserDeclaredMoveAssignment() && !hasUserDeclaredDestructor() &&
+ // In HLSL, only built-in records like resources classes can have
+ // constructors.
+ (!getLangOpts().HLSL || isLambda() ||
+ hasUserProvidedSpecialMembers());
}
/// Determine whether we need to eagerly declare a defaulted move
@@ -923,7 +951,11 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// assignment operator to be lazily declared.
bool needsImplicitCopyAssignment() const {
- return !(data().DeclaredSpecialMembers & SMF_CopyAssignment);
+ return !(data().DeclaredSpecialMembers & SMF_CopyAssignment) &&
+ // In HLSL, only built-in records like resources classes can have
+ // constructors.
+ (!getLangOpts().HLSL || isLambda() ||
+ hasUserProvidedSpecialMembers());
}
/// Determine whether we need to eagerly declare a defaulted copy
@@ -984,9 +1016,13 @@ class CXXRecordDecl : public RecordDecl {
return !(data().DeclaredSpecialMembers & SMF_MoveAssignment) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveConstructor() &&
- !hasUserDeclaredDestructor() &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable());
+ !hasUserDeclaredMoveConstructor() && !hasUserDeclaredDestructor() &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable()) &&
+ // In HLSL, only built-in records like resources classes can have
+ // constructors.
+ (!getLangOpts().HLSL ||
+ (isLambda() && lambdaIsDefaultConstructibleAndAssignable()) ||
+ hasUserProvidedSpecialMembers());
}
/// Determine whether we need to eagerly declare a move assignment
diff --git a/clang/lib/AST/Expr.cpp b/clang/lib/AST/Expr.cpp
index 64d61dbc3d128..9824367664a1e 100644
--- a/clang/lib/AST/Expr.cpp
+++ b/clang/lib/AST/Expr.cpp
@@ -2082,7 +2082,8 @@ ImplicitCastExpr *ImplicitCastExpr::Create(const ASTContext &C, QualType T,
// Per C++ [conv.lval]p3, lvalue-to-rvalue conversions on class and
// std::nullptr_t have special semantics not captured by CK_LValueToRValue.
assert((Kind != CK_LValueToRValue ||
- !(T->isNullPtrType() || T->getAsCXXRecordDecl())) &&
+ !(T->isNullPtrType() ||
+ (T->getAsCXXRecordDecl() && !C.getLangOpts().HLSL))) &&
"invalid type for lvalue-to-rvalue conversion");
ImplicitCastExpr *E =
new (Buffer) ImplicitCastExpr(T, Kind, Operand, PathSize, FPO, VK);
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index 419b3c477e7b2..28a14072982e8 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -1525,10 +1525,14 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
// isConstantInitializer produces wrong answers for structs with
// reference or bitfield members, and a few other cases, and checking
// for POD-ness protects us from some of these.
+ QualType BaseTy = getContext().getBaseElementType(Ty);
if (D.getInit() && (Ty->isArrayType() || Ty->isRecordType()) &&
(D.isConstexpr() ||
- ((Ty.isPODType(getContext()) ||
- getContext().getBaseElementType(Ty)->isObjCObjectPointerType()) &&
+ ((Ty.isPODType(getContext()) || BaseTy->isObjCObjectPointerType() ||
+ // If HLSL, check if it's a constant initializer anyway because
+ // POD-ness will no longer be true for user defined structs
+ // (since they cannot have constructors or a destructor).
+ (getLangOpts().HLSL && BaseTy->isRecordType())) &&
D.getInit()->isConstantInitializer(getContext())))) {
// If the variable's a const type, and it's neither an NRVO
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.cpp b/clang/lib/CodeGen/CGHLSLRuntime.cpp
index a134f6aab9490..575592dc61fec 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.cpp
+++ b/clang/lib/CodeGen/CGHLSLRuntime.cpp
@@ -1079,8 +1079,10 @@ void CGHLSLRuntime::emitEntryFunction(const FunctionDecl *FD,
PD->getAttr<HLSLParamModifierAttr>()) {
llvm_unreachable("Not handled yet");
} else {
- llvm::Type *ParamType =
- Param.hasByValAttr() ? Param.getParamByValType() : Param.getType();
+ llvm::Type *ParamType = Param.hasByValAttr() ? Param.getParamByValType()
+ : PD->getType()->isRecordType()
+ ? CGM.getTypes().ConvertType(PD->getType())
+ : Param.getType();
auto AttrBegin = PD->specific_attr_begin<HLSLAppliedSemanticAttr>();
auto AttrEnd = PD->specific_attr_end<HLSLAppliedSemanticAttr>();
auto Result =
@@ -1088,12 +1090,11 @@ void CGHLSLRuntime::emitEntryFunction(const FunctionDecl *FD,
SemanticValue = Result.first;
if (!SemanticValue)
return;
- if (Param.hasByValAttr()) {
+ if (Param.hasByValAttr() || PD->getType()->isRecordType()) {
llvm::Value *Var =
CGM.getLangOpts().EmitLogicalPointer
- ? cast<Instruction>(
- B.CreateStructuredAlloca(Param.getParamByValType()))
- : cast<Instruction>(B.CreateAlloca(Param.getParamByValType()));
+ ? cast<Instruction>(B.CreateStructuredAlloca(ParamType))
+ : cast<Instruction>(B.CreateAlloca(ParamType));
B.CreateStore(SemanticValue, Var);
SemanticValue = Var;
}
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index c494669420282..cba1f4e9fc6ab 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -664,8 +664,9 @@ ExprResult Sema::DefaultLvalueConversion(Expr *E) {
// We don't want to throw lvalue-to-rvalue casts on top of
// expressions of certain types in C++.
+ // In HLSL LvaluetoRvalue conversion is allowed on records.
if (getLangOpts().CPlusPlus) {
- if (T == Context.OverloadTy || T->isRecordType() ||
+ if (T == Context.OverloadTy || (T->isRecordType() && !getLangOpts().HLSL) ||
(T->isDependentType() && !T->isAnyPointerType() &&
!T->isMemberPointerType()))
return E;
@@ -16005,6 +16006,9 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
if (!LHSExpr || !RHSExpr)
return ExprError();
+ const Type *LHSTy = LHSExpr->getType().getTypePtr();
+ const Type *RHSTy = RHSExpr->getType().getTypePtr();
+
// We want to end up calling one of SemaPseudoObject::checkAssignment
// (if the LHS is a pseudo-object), BuildOverloadedBinOp (if
// both expressions are overloadable or either is type-dependent),
@@ -16012,7 +16016,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
// any placeholder types out of the way.
// Handle pseudo-objects in the LHS.
- if (const BuiltinType *pty = LHSExpr->getType()->getAsPlaceholderType()) {
+ if (const BuiltinType *pty = LHSTy->getAsPlaceholderType()) {
// Assignments with a pseudo-object l-value need special analysis.
if (pty->getKind() == BuiltinType::PseudoObject &&
BinaryOperator::isAssignmentOp(Opc))
@@ -16068,7 +16072,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
if (Opc == BO_Assign && pty->getKind() == BuiltinType::Overload) {
if (getLangOpts().CPlusPlus &&
(LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent() ||
- LHSExpr->getType()->isOverloadableType()))
+ LHSTy->isOverloadableType()))
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
return CreateBuiltinBinOp(OpLoc, Opc, LHSExpr, RHSExpr,
@@ -16077,7 +16081,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
// Don't resolve overloads if the other type is overloadable.
if (getLangOpts().CPlusPlus && pty->getKind() == BuiltinType::Overload &&
- LHSExpr->getType()->isOverloadableType())
+ LHSTy->isOverloadableType())
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
ExprResult resolvedRHS = CheckPlaceholderExpr(RHSExpr);
@@ -16085,8 +16089,8 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
RHSExpr = resolvedRHS.get();
}
- if (getLangOpts().HLSL && (LHSExpr->getType()->isHLSLResourceRecord() ||
- LHSExpr->getType()->isHLSLResourceRecordArray())) {
+ if (getLangOpts().HLSL &&
+ (LHSTy->isHLSLResourceRecord() || LHSTy->isHLSLResourceRecordArray())) {
if (!HLSL().CheckResourceBinOp(Opc, LHSExpr, RHSExpr, OpLoc))
return ExprError();
}
@@ -16094,9 +16098,17 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
if (getLangOpts().CPlusPlus) {
// Otherwise, build an overloaded op if either expression is type-dependent
// or has an overloadable type.
+ // In HLSL, user-defined structs/classes do not have ctors, dtors or
+ // overloadable operators.
+ bool IsLHSNonOverloadableHLSLType =
+ getLangOpts().HLSL && LHSTy->isRecordType() &&
+ !LHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers();
+ bool IsRHSNonOverloadableHLSLType =
+ getLangOpts().HLSL && RHSTy->isRecordType() &&
+ !RHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers();
if (LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent() ||
- LHSExpr->getType()->isOverloadableType() ||
- RHSExpr->getType()->isOverloadableType())
+ (LHSTy->isOverloadableType() && !IsLHSNonOverloadableHLSLType) ||
+ (RHSTy->isOverloadableType() && !IsRHSNonOverloadableHLSLType))
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
}
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index aba1c5072a5fc..688bb2e39cea8 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -5554,14 +5554,25 @@ bool SemaHLSL::ActOnUninitializedVarDecl(VarDecl *VD) {
if (VD->getType().getAddressSpace() == LangAS::hlsl_constant)
return true;
- // Initialize non-static resources at the global scope.
if (VD->hasGlobalStorage() && VD->getStorageClass() != SC_Static) {
const Type *Ty = VD->getType().getTypePtr();
- if (Ty->isHLSLResourceRecord())
- return initGlobalResourceDecl(VD);
- if (Ty->isHLSLResourceRecordArray())
- return initGlobalResourceArrayDecl(VD);
+ if (Ty->isHLSLResourceRecord() && initGlobalResourceDecl(VD))
+ return true;
+ if (Ty->isHLSLResourceRecordArray() && initGlobalResourceArrayDecl(VD))
+ return true;
}
+
+ // User-defined structs/classes do not have constructors.
+ // When declared at the a scope, they are part of the constant buffer
+ // and should not be initialized by the compiler.
+ // When declated at a local scope, they are by default not initialized.
+ // Also applies to arrays of user-defined structs/classes.
+ const Type *Ty = VD->getType()->getUnqualifiedDesugaredType();
+ while (Ty->isArrayType())
+ Ty = Ty->getArrayElementTypeNoTypeQual()->getUnqualifiedDesugaredType();
+ if (CXXRecordDecl *RD = Ty->getAsCXXRecordDecl())
+ return !RD->hasUserProvidedSpecialMembers();
+
return false;
}
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index e54a25405c816..2728973d38ffc 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -6927,7 +6927,11 @@ void InitializationSequence::InitializeFrom(Sema &S,
assert(S.getLangOpts().CPlusPlus);
// - If the destination type is a (possibly cv-qualified) class type:
- if (DestType->isRecordType()) {
+ // (except for HLSL, where user-defined record types do not have
+ // constructors or conversion functions)
+ if (DestType->isRecordType() &&
+ (!S.getLangOpts().HLSL ||
+ DestType->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers())) {
// - If the initialization is direct-initialization, or if it is
// copy-initialization where the cv-unqualified version of the
// source type is the same class as, or a derived class of, the
@@ -7012,7 +7016,11 @@ void InitializationSequence::InitializeFrom(Sema &S,
// - Otherwise, if the source type is a (possibly cv-qualified) class
// type, conversion functions are considered.
- if (!SourceType.isNull() && SourceType->isRecordType()) {
+ // (except for HLSL, where user-defined record types do not have
+ // constructors or conversion functions).
+ if (!SourceType.isNull() && SourceType->isRecordType() &&
+ (!S.getLangOpts().HLSL ||
+ SourceType->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers())) {
assert(Initializer && "Initializer must be non-null");
// For a conversion to _Atomic(T) from either T or a class type derived
// from T, initialize the T object then convert to _Atomic type.
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 96c4ce489fe04..f93313392e3e7 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -1822,6 +1822,9 @@ TryImplicitConversion(Sema &S, Expr *From, QualType ToType,
// given Conversion rank, in spite of the fact that a copy/move
// constructor (i.e., a user-defined conversion function) is
// called for those cases.
+ // HLSL:
+ // A conversion of an expression of class type to the same class
+ // type needs implicit LvaluetoRvalue conversion.
QualType FromType = From->getType();
if (ToType->isRecordType() &&
(S.Context.hasSameUnqualifiedType(FromType, ToType) ||
@@ -1837,6 +1840,9 @@ TryImplicitConversion(Sema &S, Expr *From, QualType ToType,
// appropriate constructor to copy the returned object, if needed.
ICS.Standard.CopyConstructor = nullptr;
+ if (S.getLangOpts().HLSL)
+ ICS.Standard.First = ICK_Lvalue_To_Rvalue;
+
// Determine whether this is considered a derived-to-base conversion.
if (!S.Context.hasSameUnqualifiedType(FromType, ToType))
ICS.Standard.Second = ICK_Derived_To_Base;
@@ -15517,7 +15523,13 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
// various built-in candidates, but as DR507 points out, this can lead to
// problems. So we do it this way, which pretty much follows what GCC does.
// Note that we go the traditional code path for compound assignment forms.
- if (Opc == BO_Assign && !Args[0]->getType()->isOverloadableType())
+ // In HLSL, user-defined structs/classes do not have ctors, dtors or
+ // overloadable operators, so we can take this shortcut too.
+ const Type *LHSTy = Args[0]->getType().getTypePtr();
+ if (Opc == BO_Assign &&
+ (!LHSTy->isOverloadableType() ||
+ (getLangOpts().HLSL && LHSTy->isRecordType() &&
+ !LHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers())))
return CreateBuiltinBinOp(OpLoc, Opc, Args[0], Args[1]);
// Build the overload set.
diff --git a/clang/test/AST/HLSL/cbuffer.hlsl b/clang/test/AST/HLSL/cbuffer.hlsl
index 487261af19133..0332600a073e1 100644
--- a/clang/test/AST/HLSL/cbuffer.hlsl
+++ b/clang/test/AST/HLSL/cbuffer.hlsl
@@ -151,7 +151,7 @@ cbuffer CB {
void f() {}
// CHECK: VarDecl {{.*}} SV 'hlsl_private float' static
static float SV;
- // CHECK: VarDecl {{.*}} s7 'EmptyStruct' callinit
+ // CHECK: VarDecl {{.*}} s7 'EmptyStruct'
EmptyStruct s7;
// CHECK: VarDecl {{.*}} Buf 'RWBuffer<float>':'hlsl::RWBuffer<float>' callinit
RWBuffer<float> Buf;
diff --git a/clang/test/AST/HLSL/matrix-constructors.hlsl b/clang/test/AST/HLSL/matrix-constructors.hlsl
index ae61ab49f8573..a00ca1a6e392e 100644
--- a/clang/test/AST/HLSL/matrix-constructors.hlsl
+++ b/clang/test/AST/HLSL/matrix-constructors.hlsl
@@ -335,6 +335,7 @@ float2x3 G = float2x3(float2x2(1,2,3,4), 5, 6);
// CHECK: VarDecl 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}, col:{{[0-9]+}}> col:{{[0-9]+}} N 'float4x4':'matrix<float, 4, 4>' cinit
// CHECK-NEXT: CXXFunctionalCastExpr 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}, col:{{[0-9]+}}> 'float4x4':'matrix<float, 4, 4>' functional cast to float4x4 <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}> 'sF' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}> 'sF' lvalue Var 0x{{[0-9a-fA-F]+}} 'f' 'sF'
struct sF {
float f[16];
@@ -385,6 +386,7 @@ float2x2 GettingStrange2 = float2x2(s3, s3);
// CHECK: VarDecl 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}, col:{{[0-9]+}}> col:{{[0-9]+}} GettingStrange3 'float2x2':'matrix<float, 2, 2>' cinit
// CHECK-NEXT: CXXFunctionalCastExpr 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}, col:{{[0-9]+}}> 'float2x2':'matrix<float, 2, 2>' functional cast to float2x2 <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}> 'S4' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr 0x{{[0-9a-fA-F]+}} <col:{{[0-9]+}}> 'S4' lvalue Var 0x{{[0-9a-fA-F]+}} 's4' 'S4'
struct S4 { float4 f;};
S4 s4;
diff --git a/clang/test/AST/HLSL/semantic-output-struct-shadow.hlsl b/clang/test/AST/HLSL/semantic-output-struct-shadow.hlsl
index e83901bb17943..def7e12785209 100644
--- a/clang/test/AST/HLSL/semantic-output-struct-shadow.hlsl
+++ b/clang/test/AST/HLSL/semantic-output-struct-shadow.hlsl
@@ -3,9 +3,9 @@
// CHECK: CXXRecordDecl {{.*}} referenced struct S definition
-// CHECK: FieldDecl {{.*}} referenced field1 'int'
+// CHECK: FieldDecl {{.*}} field1 'int'
// CHECK-NEXT: HLSLParsedSemanticAttr {{.*}} "A" 0
-// CHECK: FieldDecl {{.*}} referenced field2 'int'
+// CHECK: FieldDecl {{.*}} field2 'int'
// CHECK-NEXT: HLSLParsedSemanticAttr {{.*}} "B" 4
struct S {
diff --git a/clang/test/AST/HLSL/semantic-output-struct.hlsl b/clang/test/AST/HLSL/semantic-output-struct.hlsl
index 727c0f3040641..f43492bae28ff 100644
--- a/clang/test/AST/HLSL/semantic-output-struct.hlsl
+++ b/clang/test/AST/HLSL/semantic-output-struct.hlsl
@@ -3,9 +3,9 @@
// CHECK: CXXRecordDecl {{.*}} referenced struct S definition
-// CHECK: FieldDecl {{.*}} referenced field1 'int'
+// CHECK: FieldDecl {{.*}} field1 'int'
// CHECK-NEXT: HLSLParsedSemanticAttr {{.*}} "A" 0
-// CHECK: FieldDecl {{.*}} referenced field2 'int'
+// CHECK: FieldDecl {{.*}} field2 'int'
// CHECK-NEXT: HLSLParsedSemanticAttr {{.*}} "B" 4
struct S {
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
index 740b80afdb609..8d5ed58e58d74 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
@@ -157,8 +157,8 @@ struct Derived : BFields {
// flatten from a derived struct with bitfields
// CHECK-LABEL: call8
// CHECK: [[A:%.*]] = alloca [4 x i32], align 4
-// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
+// CHECK: [[Tmp:%.*]] = alloca %struct.Derived, align 1
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 0
// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 1
// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 2
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl
index e9661a6e2b8be..3be7d791a9fe5 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl
@@ -66,6 +66,8 @@ struct UnnamedDerived : UnnamedOnly {};
// CHECK-LABEL: define hidden void @_Z5case1v(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_RESULT]], ptr align 1 @__const._Z5case1v.TF1, i32 8, i1 false)
// CHECK-NEXT: ret void
//
@@ -78,6 +80,8 @@ TwoFloats case1() {
// CHECK-LABEL: define hidden void @_Z5case2v(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_RESULT]], ptr align 1 @__const._Z5case2v.TF2, i32 8, i1 false)
// CHECK-NEXT: ret void
//
@@ -90,7 +94,9 @@ TwoFloats case2() {
// CHECK-LABEL: define hidden void @_Z5case3i(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], i32 noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca i32, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store i32 [[VAL]], ptr [[VAL_ADDR]], align 4
// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[VAL_ADDR]], align 4
@@ -110,7 +116,9 @@ TwoFloats case3(int Val) {
// CHECK-LABEL: define hidden void @_Z5case4Dv2_i(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], <2 x i32> noundef [[TWOVALS:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TWOVALS_ADDR:%.*]] = alloca <2 x i32>, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store <2 x i32> [[TWOVALS]], ptr [[TWOVALS_ADDR]], align 4
// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load <2 x i32>, ptr [[TWOVALS_ADDR]], align 4
@@ -133,7 +141,9 @@ TwoFloats case4(int2 TwoVals) {
// CHECK-LABEL: define hidden void @_Z5case5Dv2_i(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOINTS:%.*]]) align 1 [[AGG_RESULT:%.*]], <2 x i32> noundef [[TWOVALS:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TWOVALS_ADDR:%.*]] = alloca <2 x i32>, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store <2 x i32> [[TWOVALS]], ptr [[TWOVALS_ADDR]], align 4
// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load <2 x i32>, ptr [[TWOVALS_ADDR]], align 4
@@ -153,10 +163,14 @@ TwoInts case5(int2 TwoVals) {
// Case 6: Initialization from a scalarized structure of different type with
// different element types.
// CHECK-LABEL: define hidden void @_Z5case69TwoFloats(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOINTS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS:%.*]]) align 1 [[TF4:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOINTS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[TF4:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TF4_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[TF4]], ptr [[TF4_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[AGG_RESULT]], i32 0, i32 0
-// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[TF4]], i32 0, i32 0
+// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS:%.*]], ptr [[TF4]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[X]], align 1
// CHECK-NEXT: [[CONV:%.*]] = fptosi float [[TMP0]] to i32
// CHECK-NEXT: store i32 [[CONV]], ptr [[Z]], align 1
@@ -175,12 +189,26 @@ TwoInts case6(TwoFloats TF4) {
// Case 7: Initialization of a complex structure, with bogus braces and element
// conversions from a collection of scalar values, and structures.
// CHECK-LABEL: define hidden void @_Z5case77TwoIntsS_i9TwoFloatsS0_S0_S0_(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_DOGGO:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_TWOINTS:%.*]]) align 1 [[TI1:%.*]], ptr noundef byval([[STRUCT_TWOINTS]]) align 1 [[TI2:%.*]], i32 noundef [[VAL:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS:%.*]]) align 1 [[TF1:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS]]) align 1 [[TF2:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS]]) align 1 [[TF3:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS]]) align 1 [[TF4:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_DOGGO:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[TI1:%.*]], ptr noundef dead_on_return [[TI2:%.*]], i32 noundef [[VAL:%.*]], ptr noundef dead_on_return [[TF1:%.*]], ptr noundef dead_on_return [[TF2:%.*]], ptr noundef dead_on_return [[TF3:%.*]], ptr noundef dead_on_return [[TF4:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TI1_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TI2_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca i32, align 4
+// CHECK-NEXT: [[TF1_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TF2_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TF3_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TF4_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[TI1]], ptr [[TI1_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[TI2]], ptr [[TI2_INDIRECT_ADDR]], align 4
// CHECK-NEXT: store i32 [[VAL]], ptr [[VAL_ADDR]], align 4
+// CHECK-NEXT: store ptr [[TF1]], ptr [[TF1_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[TF2]], ptr [[TF2_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[TF3]], ptr [[TF3_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[TF4]], ptr [[TF4_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[LEGSTATE:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO]], ptr [[AGG_RESULT]], i32 0, i32 0
-// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI1]], i32 0, i32 0
+// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS:%.*]], ptr [[TI1]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[Z]], align 1
// CHECK-NEXT: [[VECINIT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i32 0
// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI1]], i32 0, i32 1
@@ -201,7 +229,7 @@ TwoInts case6(TwoFloats TF4) {
// CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP5]] to float
// CHECK-NEXT: store float [[CONV]], ptr [[HAIRCOUNT]], align 1
// CHECK-NEXT: [[EARDIRECTION:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO]], ptr [[AGG_RESULT]], i32 0, i32 3
-// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[TF1]], i32 0, i32 0
+// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS:%.*]], ptr [[TF1]], i32 0, i32 0
// CHECK-NEXT: [[TMP6:%.*]] = load float, ptr [[X]], align 1
// CHECK-NEXT: [[VECINIT6:%.*]] = insertelement <4 x float> poison, float [[TMP6]], i32 0
// CHECK-NEXT: [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[TF1]], i32 0, i32 1
@@ -239,10 +267,14 @@ Doggo case7(TwoInts TI1, TwoInts TI2, int Val, TwoFloats TF1, TwoFloats TF2,
// Case 8: Initialization of a structure from a different structure with
// significantly different element types and grouping.
// CHECK-LABEL: define hidden void @_Z5case85Doggo(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_ANIMALBITS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_DOGGO:%.*]]) align 1 [[D1:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_ANIMALBITS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[D1:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[D1_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[D1]], ptr [[D1_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[LEGS:%.*]] = getelementptr inbounds nuw [[STRUCT_ANIMALBITS]], ptr [[AGG_RESULT]], i32 0, i32 0
-// CHECK-NEXT: [[LEGSTATE:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO]], ptr [[D1]], i32 0, i32 0
+// CHECK-NEXT: [[LEGSTATE:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO:%.*]], ptr [[D1]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[LEGSTATE]], align 1
// CHECK-NEXT: [[VECEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i64 0
// CHECK-NEXT: store i32 [[VECEXT]], ptr [[LEGS]], align 1
@@ -325,10 +357,16 @@ AnimalBits case8(Doggo D1) {
// structures from different layouts, different component groupings, with no
// top-level bracing separation.
// CHECK-LABEL: define hidden void @_Z5case95Doggo10AnimalBits(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_ZOO:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_DOGGO:%.*]]) align 1 [[D1:%.*]], ptr noundef byval([[STRUCT_ANIMALBITS:%.*]]) align 1 [[A1:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_ZOO:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[D1:%.*]], ptr noundef dead_on_return [[A1:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[D1_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[A1_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[D1]], ptr [[D1_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[A1]], ptr [[A1_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[DOGS:%.*]] = getelementptr inbounds nuw [[STRUCT_ZOO]], ptr [[AGG_RESULT]], i32 0, i32 0
-// CHECK-NEXT: [[LEGSTATE:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO]], ptr [[DOGS]], i32 0, i32 0
+// CHECK-NEXT: [[LEGSTATE:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO:%.*]], ptr [[DOGS]], i32 0, i32 0
// CHECK-NEXT: [[LEGSTATE1:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO]], ptr [[D1]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[LEGSTATE1]], align 1
// CHECK-NEXT: [[VECEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i64 0
@@ -400,7 +438,7 @@ AnimalBits case8(Doggo D1) {
// CHECK-NEXT: store <4 x float> [[VECINIT43]], ptr [[ARRAYINIT_ELEMENT]], align 1
// CHECK-NEXT: [[ARRAYINIT_ELEMENT44:%.*]] = getelementptr inbounds [[STRUCT_DOGGO]], ptr [[DOGS]], i32 1
// CHECK-NEXT: [[LEGSTATE45:%.*]] = getelementptr inbounds nuw [[STRUCT_DOGGO]], ptr [[ARRAYINIT_ELEMENT44]], i32 0, i32 0
-// CHECK-NEXT: [[LEGS:%.*]] = getelementptr inbounds nuw [[STRUCT_ANIMALBITS]], ptr [[A1]], i32 0, i32 0
+// CHECK-NEXT: [[LEGS:%.*]] = getelementptr inbounds nuw [[STRUCT_ANIMALBITS:%.*]], ptr [[A1]], i32 0, i32 0
// CHECK-NEXT: [[ARRAYIDX46:%.*]] = getelementptr inbounds nuw [4 x i32], ptr [[LEGS]], i32 0, i32 0
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX46]], align 1
// CHECK-NEXT: [[VECINIT47:%.*]] = insertelement <4 x i32> poison, i32 [[TMP14]], i32 0
@@ -741,9 +779,15 @@ Zoo case9(Doggo D1, AnimalBits A1) {
// Case 10: Initialize an object with a base class from two objects.
// CHECK-LABEL: define hidden void @_Z6case109TwoFloatsS_(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_FOURFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS:%.*]]) align 1 [[TF1:%.*]], ptr noundef byval([[STRUCT_TWOFLOATS]]) align 1 [[TF2:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_FOURFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[TF1:%.*]], ptr noundef dead_on_return [[TF2:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 0
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TF1_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TF2_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[TF1]], ptr [[TF1_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[TF2]], ptr [[TF2_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS:%.*]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[X1:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[TF1]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[X1]], align 1
// CHECK-NEXT: store float [[TMP0]], ptr [[X]], align 1
@@ -770,11 +814,13 @@ FourFloats case10(TwoFloats TF1, TwoFloats TF2) {
// CHECK-LABEL: define hidden void @_Z6case11f(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_FOURFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], float noundef nofpclass(nan inf) [[F:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[F_ADDR:%.*]] = alloca float, align 4
// CHECK-NEXT: [[REF_TMP:%.*]] = alloca <4 x float>, align 4
// CHECK-NEXT: [[REF_TMP1:%.*]] = alloca <4 x float>, align 4
// CHECK-NEXT: [[REF_TMP4:%.*]] = alloca <4 x float>, align 4
// CHECK-NEXT: [[REF_TMP7:%.*]] = alloca <4 x float>, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store float [[F]], ptr [[F_ADDR]], align 4
// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS:%.*]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F_ADDR]], align 4
@@ -819,8 +865,10 @@ FourFloats case11(float F) {
// CHECK-LABEL: define hidden void @_Z6case12ii(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_SLICYBITS:%.*]]) align 1 [[AGG_RESULT:%.*]], i32 noundef [[I:%.*]], i32 noundef [[J:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[I_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[J_ADDR:%.*]] = alloca i32, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store i32 [[I]], ptr [[I_ADDR]], align 4
// CHECK-NEXT: store i32 [[J]], ptr [[J_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I_ADDR]], align 4
@@ -839,9 +887,13 @@ SlicyBits case12(int I, int J) {
// Case 13: Initialize bitfield from a struct of two ints.
// CHECK-LABEL: define hidden void @_Z6case137TwoInts(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_SLICYBITS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_TWOINTS:%.*]]) align 1 [[TI:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_SLICYBITS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[TI:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 0
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[TI_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[TI]], ptr [[TI_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS:%.*]], ptr [[TI]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[Z]], align 1
// CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[TMP0]] to i8
// CHECK-NEXT: store i8 [[TMP1]], ptr [[AGG_RESULT]], align 1
@@ -859,14 +911,18 @@ SlicyBits case13(TwoInts TI) {
// Case 14: Initialize struct of ints from struct with bitfields.
// CHECK-LABEL: define hidden void @_Z6case149SlicyBits(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOINTS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_SLICYBITS:%.*]]) align 1 [[SB:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOINTS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[SB:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[SB_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[SB]], ptr [[SB_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[BF_LOAD:%.*]] = load i8, ptr [[SB]], align 1
// CHECK-NEXT: [[BF_CAST:%.*]] = sext i8 [[BF_LOAD]] to i32
// CHECK-NEXT: store i32 [[BF_CAST]], ptr [[Z]], align 1
// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[AGG_RESULT]], i32 0, i32 1
-// CHECK-NEXT: [[W1:%.*]] = getelementptr inbounds nuw [[STRUCT_SLICYBITS]], ptr [[SB]], i32 0, i32 1
+// CHECK-NEXT: [[W1:%.*]] = getelementptr inbounds nuw [[STRUCT_SLICYBITS:%.*]], ptr [[SB]], i32 0, i32 1
// CHECK-NEXT: [[BF_LOAD2:%.*]] = load i8, ptr [[W1]], align 1
// CHECK-NEXT: [[BF_CAST3:%.*]] = sext i8 [[BF_LOAD2]] to i32
// CHECK-NEXT: store i32 [[BF_CAST3]], ptr [[W]], align 1
@@ -879,15 +935,19 @@ TwoInts case14(SlicyBits SB) {
// Case 15: Initialize struct of floats from struct with bitfields.
// CHECK-LABEL: define hidden void @_Z6case159SlicyBits(
-// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_SLICYBITS:%.*]]) align 1 [[SB:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noundef dead_on_return [[SB:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[SB_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
+// CHECK-NEXT: store ptr [[SB]], ptr [[SB_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[BF_LOAD:%.*]] = load i8, ptr [[SB]], align 1
// CHECK-NEXT: [[BF_CAST:%.*]] = sext i8 [[BF_LOAD]] to i32
// CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[BF_CAST]] to float
// CHECK-NEXT: store float [[CONV]], ptr [[X]], align 1
// CHECK-NEXT: [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 1
-// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_SLICYBITS]], ptr [[SB]], i32 0, i32 1
+// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_SLICYBITS:%.*]], ptr [[SB]], i32 0, i32 1
// CHECK-NEXT: [[BF_LOAD1:%.*]] = load i8, ptr [[W]], align 1
// CHECK-NEXT: [[BF_CAST2:%.*]] = sext i8 [[BF_LOAD1]] to i32
// CHECK-NEXT: [[CONV3:%.*]] = sitofp i32 [[BF_CAST2]] to float
@@ -904,7 +964,9 @@ TwoFloats case15(SlicyBits SB) {
// CHECK-LABEL: define hidden void @_Z7makeTwoRf(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_TWOFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]], ptr noalias noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[X_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store ptr [[X]], ptr [[X_ADDR]], align 4
// CHECK-NEXT: [[X1:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4, !nonnull [[META3:![0-9]+]], !align [[META4:![0-9]+]]
@@ -930,9 +992,11 @@ TwoFloats makeTwo(inout float X) {
// CHECK-LABEL: define hidden void @_Z6case16v(
// CHECK-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_FOURFLOATS:%.*]]) align 1 [[AGG_RESULT:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RESULT_PTR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[X:%.*]] = alloca float, align 4
// CHECK-NEXT: [[REF_TMP:%.*]] = alloca [[STRUCT_TWOFLOATS:%.*]], align 1
// CHECK-NEXT: [[TMP:%.*]] = alloca float, align 4
+// CHECK-NEXT: store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// CHECK-NEXT: store float 0.000000e+00, ptr [[X]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[X]], align 4
// CHECK-NEXT: store float [[TMP0]], ptr [[TMP]], align 4
@@ -1002,11 +1066,13 @@ void case18() {
// InitList with Struct with unnamed bitfield on RHS
// CHECK-LABEL: define hidden void @_Z6case197Unnamed(
-// CHECK-SAME: ptr noundef byval([[STRUCT_UNNAMED:%.*]]) align 1 [[U:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[U:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[U_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1
+// CHECK-NEXT: store ptr [[U]], ptr [[U_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 0
-// CHECK-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_UNNAMED]], ptr [[U]], i32 0, i32 0
+// CHECK-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_UNNAMED:%.*]], ptr [[U]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A]], align 1
// CHECK-NEXT: store i32 [[TMP0]], ptr [[Z]], align 1
// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 1
@@ -1031,9 +1097,11 @@ void case20() {
// InitList with Empty Struct on RHS
// CHECK-LABEL: define hidden void @_Z6case215Empty(
-// CHECK-SAME: ptr noundef byval([[STRUCT_EMPTY:%.*]]) align 1 [[E:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[E:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[E_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1
+// CHECK-NEXT: store ptr [[E]], ptr [[E_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI]], ptr align 1 @__const._Z6case215Empty.TI, i32 8, i1 false)
// CHECK-NEXT: ret void
//
@@ -1055,9 +1123,11 @@ void case22() {
// InitList with Struct with only unnamed bitfield on RHS
// CHECK-LABEL: define hidden void @_Z6case2311UnnamedOnly(
-// CHECK-SAME: ptr noundef byval([[STRUCT_UNNAMEDONLY:%.*]]) align 1 [[UO:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[UO:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[UO_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1
+// CHECK-NEXT: store ptr [[UO]], ptr [[UO_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI]], ptr align 1 @__const._Z6case2311UnnamedOnly.TI, i32 8, i1 false)
// CHECK-NEXT: ret void
//
@@ -1082,10 +1152,14 @@ void case24() {
}
// CHECK-LABEL: define hidden void @_Z6case2512EmptyDerived14UnnamedDerived(
-// CHECK-SAME: ptr noundef byval([[STRUCT_EMPTYDERIVED:%.*]]) align 1 [[ED:%.*]], ptr noundef byval([[STRUCT_UNNAMEDDERIVED:%.*]]) align 1 [[UD:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[ED:%.*]], ptr noundef dead_on_return [[UD:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[ED_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[UD_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TI1:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1
// CHECK-NEXT: [[TI2:%.*]] = alloca [[STRUCT_TWOINTS]], align 1
+// CHECK-NEXT: store ptr [[ED]], ptr [[ED_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: store ptr [[UD]], ptr [[UD_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI1]], ptr align 1 @__const._Z6case2512EmptyDerived14UnnamedDerived.TI1, i32 8, i1 false)
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI2]], ptr align 1 @__const._Z6case2512EmptyDerived14UnnamedDerived.TI2, i32 8, i1 false)
// CHECK-NEXT: ret void
@@ -1096,11 +1170,13 @@ void case25(EmptyDerived ED, UnnamedDerived UD) {
}
// CHECK-LABEL: define hidden void @_Z6case267TwoInts(
-// CHECK-SAME: ptr noundef byval([[STRUCT_TWOINTS:%.*]]) align 1 [[TI:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[TI:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[TI_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[F:%.*]] = alloca <4 x float>, align 4
// CHECK-NEXT: [[F2:%.*]] = alloca <3 x float>, align 4
-// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 0
+// CHECK-NEXT: store ptr [[TI]], ptr [[TI_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS:%.*]], ptr [[TI]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[Z]], align 1
// CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
// CHECK-NEXT: [[VECINIT:%.*]] = insertelement <4 x float> poison, float [[CONV]], i32 0
@@ -1134,9 +1210,11 @@ struct CustomResource {
};
// CHECK-LABEL: define hidden void @_Z6case2714CustomResource(
-// CHECK-SAME: ptr noundef byval([[STRUCT_CUSTOMRESOURCE:%.*]]) align 1 [[A:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[A:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[B:%.*]] = alloca [[STRUCT_CUSTOMRESOURCE]], align 1
+// CHECK-NEXT: [[A_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[B:%.*]] = alloca [[STRUCT_CUSTOMRESOURCE:%.*]], align 1
+// CHECK-NEXT: store ptr [[A]], ptr [[A_INDIRECT_ADDR]], align 4
// CHECK-NEXT: [[H:%.*]] = getelementptr inbounds nuw [[STRUCT_CUSTOMRESOURCE]], ptr [[B]], i32 0, i32 0
// CHECK-NEXT: [[H1:%.*]] = getelementptr inbounds nuw [[STRUCT_CUSTOMRESOURCE]], ptr [[A]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load target("dx.TypedBuffer", float, 1, 0, 0), ptr [[H1]], align 1
@@ -1150,11 +1228,13 @@ void case27(CustomResource a) {
// Check cases with explicit casts
// CHECK-LABEL: define hidden void @_Z6case289TwoFloats(
-// CHECK-SAME: ptr noundef byval([[STRUCT_TWOFLOATS:%.*]]) align 1 [[TF:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[TF:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[TF_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1
// CHECK-NEXT: [[REF_TMP:%.*]] = alloca [[STRUCT_TWOINTS]], align 1
-// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca [[STRUCT_TWOFLOATS]], align 1
+// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca [[STRUCT_TWOFLOATS:%.*]], align 1
+// CHECK-NEXT: store ptr [[TF]], ptr [[TF_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_TEMP]], ptr align 1 [[TF]], i32 8, i1 false)
// CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds [[STRUCT_TWOINTS]], ptr [[REF_TMP]], i32 0, i32 0
// CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds [[STRUCT_TWOINTS]], ptr [[REF_TMP]], i32 0, i32 1
@@ -1172,8 +1252,8 @@ void case27(CustomResource a) {
// CHECK-NEXT: store i32 [[TMP2]], ptr [[Z]], align 1
// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 1
// CHECK-NEXT: [[W6:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[REF_TMP]], i32 0, i32 1
-// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[W6]], align 1
-// CHECK-NEXT: store i32 [[TMP5]], ptr [[W]], align 1
+// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[W6]], align 1
+// CHECK-NEXT: store i32 [[TMP3]], ptr [[W]], align 1
// CHECK-NEXT: ret void
//
void case28(TwoFloats TF) {
@@ -1181,11 +1261,13 @@ void case28(TwoFloats TF) {
}
// CHECK-LABEL: define hidden void @_Z6case2910FourFloats(
-// CHECK-SAME: ptr noundef byval([[STRUCT_FOURFLOATS:%.*]]) align 1 [[FF:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[FF:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[FF_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[INTS:%.*]] = alloca [2 x i32], align 4
// CHECK-NEXT: [[REF_TMP:%.*]] = alloca [2 x i32], align 4
-// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca [[STRUCT_FOURFLOATS]], align 1
+// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca [[STRUCT_FOURFLOATS:%.*]], align 1
+// CHECK-NEXT: store ptr [[FF]], ptr [[FF_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_TEMP]], ptr align 1 [[FF]], i32 16, i1 false)
// CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds [2 x i32], ptr [[REF_TMP]], i32 0, i32 0
// CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds [2 x i32], ptr [[REF_TMP]], i32 0, i32 1
@@ -1204,8 +1286,8 @@ void case28(TwoFloats TF) {
// CHECK-NEXT: store i32 [[TMP2]], ptr [[INTS]], align 4
// CHECK-NEXT: [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds i32, ptr [[INTS]], i32 1
// CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds nuw [2 x i32], ptr [[REF_TMP]], i32 0, i32 1
-// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[ARRAYIDX7]], align 4
-// CHECK-NEXT: store i32 [[TMP5]], ptr [[ARRAYINIT_ELEMENT]], align 4
+// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[ARRAYIDX7]], align 4
+// CHECK-NEXT: store i32 [[TMP3]], ptr [[ARRAYINIT_ELEMENT]], align 4
// CHECK-NEXT: ret void
//
void case29(FourFloats FF) {
@@ -1239,8 +1321,8 @@ void case29(FourFloats FF) {
// CHECK-NEXT: store i32 [[TMP2]], ptr [[Z]], align 1
// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 1
// CHECK-NEXT: [[W8:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[REF_TMP]], i32 0, i32 1
-// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[W8]], align 1
-// CHECK-NEXT: store i32 [[TMP5]], ptr [[W]], align 1
+// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[W8]], align 1
+// CHECK-NEXT: store i32 [[TMP3]], ptr [[W]], align 1
// CHECK-NEXT: ret void
//
void case30() {
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/MatrixElementTypeCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/MatrixElementTypeCast.hlsl
index dd9dd706aae26..a717ba5d6b2f1 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/MatrixElementTypeCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/MatrixElementTypeCast.hlsl
@@ -165,11 +165,13 @@ struct Derived : BFields {
};
// CHECK-LABEL: define hidden void @_Z5call47Derived(
-// CHECK-SAME: ptr noundef byval([[STRUCT_DERIVED:%.*]]) align 1 [[D:%.*]]) #[[ATTR0]] {
+// CHECK-SAME: ptr noundef dead_on_return [[D:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[A:%.*]] = alloca [2 x <2 x i32>], align 4
-// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca [[STRUCT_DERIVED]], align 1
+// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca [[STRUCT_DERIVED:.*]], align 1
// CHECK-NEXT: [[FLATCAST_TMP:%.*]] = alloca <4 x i32>, align 4
+// CHECK-NEXT: store ptr %D, ptr [[D_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_TEMP]], ptr align 1 [[D]], i32 19, i1 false)
// CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds [[STRUCT_DERIVED]], ptr [[AGG_TEMP]], i32 0, i32 0
// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw [[STRUCT_BFIELDS:%.*]], ptr [[GEP]], i32 0, i32 1
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl
index ab5873bfa8296..3b7111ee2fd02 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl
@@ -152,8 +152,11 @@ struct Derived : BFields {
// Derived Struct truncate to scalar
// CHECK-LABEL: call9
-// CHECK: [[D2:%.*]] = alloca double, align 8
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[D2:%.*]] = alloca double, align 8
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
+// CHECK-NEXT: store ptr %D, ptr [[D_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0
// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1
@@ -200,8 +203,11 @@ export void call10(int4 I) {
// truncate derived struct
// CHECK-LABEL: call11
-// CHECK: [[B:%.*]] = alloca %struct.BFields, align 1
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[B:%.*]] = alloca %struct.BFields, align 1
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
+// CHECK-NEXT: store ptr %D, ptr [[D_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[D]], i32 19, i1 false)
// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.BFields, ptr [[B]], i32 0
// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl
index ad227f9fe825e..31263bca443a6 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl
@@ -94,9 +94,12 @@ struct Derived : BFields {
// vector flat cast from derived struct with bitfield
// CHECK-LABEL: call6
-// CHECK: [[A:%.*]] = alloca <4 x i32>, align 4
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[A:%.*]] = alloca <4 x i32>, align 4
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
// CHECK-NEXT: [[FlatTmp:%.*]] = alloca <4 x i32>, align 4
+// CHECK-NEXT: store ptr %D, ptr [[D_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0
// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1
@@ -217,9 +220,12 @@ struct BoolVecStruct {
// vector flat cast from struct containing bool vector
// CHECK-LABEL: call10
-// CHECK: [[V:%.*]] = alloca <2 x i32>, align 4
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[V:%.*]] = alloca <2 x i32>, align 4
// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca %struct.BoolVecStruct, align 1
// CHECK-NEXT: [[FLATCAST_TMP:%.*]] = alloca <2 x i1>, align 4
+// CHECK-NEXT: store ptr %s, ptr [[S_INDIRECT_ADDR]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_TEMP]], ptr align 1 %s, i32 8, i1 false)
// CHECK-NEXT: [[VECTOR_GEP:%.*]] = getelementptr inbounds %struct.BoolVecStruct, ptr [[AGG_TEMP]], i32 0, i32 0
// CHECK-NEXT: [[TMP0:%.*]] = load <2 x i1>, ptr [[FLATCAST_TMP]], align 4
diff --git a/clang/test/CodeGenHLSL/builtins/hlsl_resource_t.hlsl b/clang/test/CodeGenHLSL/builtins/hlsl_resource_t.hlsl
index 0eb36ce8fb7bb..a98f231014797 100644
--- a/clang/test/CodeGenHLSL/builtins/hlsl_resource_t.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/hlsl_resource_t.hlsl
@@ -11,9 +11,9 @@ struct CustomResource {
// CHECK: %"class.hlsl::StructuredBuffer" = type { target("dx.RawBuffer", %struct.MyStruct, 0, 0)
// CHECK: %struct.MyStruct = type { <4 x float>, <2 x i32> }
-// CHECK: define hidden void @_Z2fa14CustomResource(ptr noundef byval(%struct.CustomResource) align 1 %a)
-// CHECK: call void @_Z4foo114CustomResource(ptr noundef byval(%struct.CustomResource) align 1 %agg.tmp)
-// CHECK: declare hidden void @_Z4foo114CustomResource(ptr noundef byval(%struct.CustomResource) align 1)
+// CHECK: define hidden void @_Z2fa14CustomResource(ptr noundef dead_on_return %a)
+// CHECK: call void @_Z4foo114CustomResource(ptr noundef dead_on_return %byval-temp)
+// CHECK: declare hidden void @_Z4foo114CustomResource(ptr noundef dead_on_return)
void foo1(CustomResource res);
@@ -21,7 +21,7 @@ void fa(CustomResource a) {
foo1(a);
}
-// CHECK: define hidden void @_Z2fb14CustomResource(ptr noundef byval(%struct.CustomResource) align 1 %a)
+// CHECK: define hidden void @_Z2fb14CustomResource(ptr noundef dead_on_return %a)
void fb(CustomResource a) {
CustomResource b = a;
}
diff --git a/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl b/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
index 2e6a7ef86c610..4e1c1b7b55984 100644
--- a/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
+++ b/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
@@ -18,6 +18,8 @@ RasterizerOrderedStructuredBuffer<int> Out2;
void main(unsigned GI : SV_GroupIndex) {
// CHECK: define void @main()
+ // CHECK: %[[TMP:.*]] = alloca %struct.S, align 1
+
// DXIL: %[[INPTR:.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_i32_0_0t.i32(target("dx.RawBuffer", i32, 0, 0) %{{.*}}, i32 %{{.*}})
// SPV: %[[INPTR:.*]] = call noundef align 4 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0i32_12_0t.i32(target("spirv.VulkanBuffer", [0 x i32], 12, 0) %{{.*}}, i32 %{{.*}})
// CHECK: %[[LOAD:.*]] = load i32, ptr {{.*}}%[[INPTR]]
@@ -34,18 +36,14 @@ void main(unsigned GI : SV_GroupIndex) {
Out2[GI] = In[GI];
#endif
- // For SPIR-V, the addrspacecast comes from `S::operator=` member function, which expects
- // parameters in address space 0. This is why hlsl_device is a sub address
- // space of the default address space.
- // SPV: %[[INPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
- // SPV: %[[INCAST:.*]] = addrspacecast ptr addrspace(11) %[[INPTR]] to ptr
// SPV: %[[OUTPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
- // SPV: %[[OUTCAST:.*]] = addrspacecast ptr addrspace(11) %[[OUTPTR]] to ptr
- // SPV: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[OUTCAST]], ptr align 1 %[[INCAST]], i64 4, i1 false)
+ // SPV: %[[INPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
+ // SPV: call void @llvm.memcpy.p11.p11.i64(ptr addrspace(11) align 1 %[[OUTPTR]], ptr addrspace(11) align 1 %[[INPTR]], i64 4, i1 false)
+ // SPV: call void @llvm.memcpy.p0.p11.i64(ptr align 1 %[[TMP]], ptr addrspace(11) align 1 %[[OUTPTR]], i64 4, i1 false)
- // For DXIL, hlsl_device and the default address space map to the same target address space. No need for an address space cast.
- // DXIL: %[[INPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
// DXIL: %[[OUTPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
+ // DXIL: %[[INPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
// DXIL: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %[[OUTPTR]], ptr align 1 %[[INPTR]], i32 4, i1 false)
+ // DXIL: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %[[TMP]], ptr align 1 %[[OUTPTR]], i32 4, i1 false)
RWSB3[0] = RWSB3[1];
}
diff --git a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
index 506de82412f8a..2f7dc02c6d7f2 100644
--- a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
+++ b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
@@ -1,38 +1,102 @@
-// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -finclude-default-header -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s --check-prefixes=CHECK-DXIL
-// RUN: %clang_cc1 -triple spirv-pc-vulkan1.3-library -finclude-default-header -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s --check-prefixes=CHECK-SPIR
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -finclude-default-header -emit-llvm -disable-llvm-passes -o - %s | \
+// RUN: FileCheck %s -DCONST_ADDR_SPACE=2 -DPADDING_TYPE="dx.Padding"
-struct S {
+// RUN: %clang_cc1 -triple spirv-pc-vulkan1.3-library -finclude-default-header -emit-llvm -disable-llvm-passes -o - %s | \
+// RUN: FileCheck %s -DCONST_ADDR_SPACE=12 -DPADDING_TYPE="spirv.Padding" --check-prefixes=CHECK,SPIRV
+
+struct P {
float3 a;
+};
+
+struct S : P {
float4 b;
};
-// CHECK-DXIL-DAG: %S = type <{ <3 x float>, target("dx.Padding", 4), <4 x float> }>
-// CHECK-DXIL-DAG: %struct.S = type { <3 x float>, <4 x float> }
-// CHECK-SPIR-DAG: %S = type <{ <3 x float>, target("spirv.Padding", 4), <4 x float> }>
-// CHECK-SPIR-DAG: %struct.S = type { <3 x float>, <4 x float> }
+
+// CHECK-DAG: %S = type <{ <3 x float>, target("[[PADDING_TYPE]]", 4), <4 x float> }>
+// CHECK-DAG: %struct.P = type { <3 x float> }
+// CHECK-DAG: %struct.S = type { %struct.P, <4 x float> }
cbuffer CB {
S cbs;
};
-// CHECK-DXIL-DAG: @cbs = external hidden addrspace(2) global %S, align 1
-// CHECK-SPIR-DAG: @cbs = external hidden addrspace(12) global %S, align 1
-
-void main() {
- S tmp = (S)cbs;
-// CHECK-DXIL: %agg-temp = alloca %struct.S, align 1
-// CHECK-DXIL: %[[#DST:]] = getelementptr inbounds %struct.S, ptr %agg-temp, i32 0, i32 0
-// CHECK-DXIL: %cbuf.load = load <3 x float>, ptr addrspace(2) @cbs, align 4
-// CHECK-DXIL: store <3 x float> %cbuf.load, ptr %[[#DST]], align 4
-
-// CHECK-DXIL: %[[#DST:]] = getelementptr inbounds %struct.S, ptr %agg-temp, i32 0, i32 1
-// CHECK-DXIL: %cbuf.load1 = load <4 x float>, ptr addrspace(2) getelementptr inbounds nuw (i8, ptr addrspace(2) @cbs, i32 16), align 4
-// CHECK-DXIL: store <4 x float> %cbuf.load1, ptr %[[#DST]], align 4
-
-// CHECK-SPIR: %agg-temp = alloca %struct.S, align 1
-// CHECK-SPIR: %[[#DST:]] = getelementptr inbounds %struct.S, ptr %agg-temp, i32 0, i32 0
-// CHECK-SPIR: %cbuf.load = load <3 x float>, ptr addrspace(12) @cbs, align 4
-// CHECK-SPIR: store <3 x float> %cbuf.load, ptr %[[#DST]], align 4
-
-// CHECK-SPIR: %[[#DST:]] = getelementptr inbounds %struct.S, ptr %agg-temp, i32 0, i32 1
-// CHECK-SPIR: %cbuf.load1 = load <4 x float>, ptr addrspace(12) getelementptr inbounds nuw (i8, ptr addrspace(12) @cbs, i64 16), align 4
-// CHECK-SPIR: store <4 x float> %cbuf.load1, ptr %[[#DST]], align 4
+// CHECK-DAG: @cbs = external hidden addrspace([[CONST_ADDR_SPACE]]) global %S, align 1
+
+// CHECK-LABEL: case1
+// CHECK-NEXT: entry:
+// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+//
+ // Copy S field by field into local variable in default address space.
+//
+// CHECK-NEXT: [[LocalS:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[PtrA:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 0
+// CHECK-NEXT: [[CBufLoad1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
+// CHECK-NEXT: store <3 x float> [[CBufLoad1]], ptr [[PtrA]], align 4
+// CHECK-NEXT: [[PtrB:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 1
+// CHECK-NEXT: [[CBufLoad2:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 4
+// CHECK-NEXT: store <4 x float> [[CBufLoad2]], ptr [[PtrB]], align 4
+// CHECK-NEXT: ret void
+void case1() {
+ S local = cbs;
+}
+
+// CHECK-LABEL: case2
+// CHECK-NEXT: entry:
+// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+//
+// Copy S field by field into a temporary variable in default address space.
+//
+// CHECK-NEXT: [[LocalS:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[AggTemp:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[PtrA:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
+// CHECK-NEXT: [[CBufLoad1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
+// CHECK-NEXT: store <3 x float> [[CBufLoad1]], ptr [[PtrA]], align 4
+// CHECK-NEXT: [[PtrB:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+// CHECK-NEXT: [[CBufLoad2:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 4
+// CHECK-NEXT: store <4 x float> [[CBufLoad2]], ptr [[PtrB]], align 4
+//
+// The proces HLSLElementwiseCast - copy individual vector elements between the structs.
+//
+// CHECK-NEXT: [[VecGep1:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 0
+// CHECK-NEXT: [[VecGep2:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 1
+// CHECK-NEXT: [[VecGep3:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
+// CHECK-NEXT: [[VecGep4:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+
+// CHECK-NEXT: [[VecA1:%.*]] = load <3 x float>, ptr [[VecGep3]], align 4
+// CHECK-NEXT: [[Val1:%.*]] = extractelement <3 x float> [[VecA1]], i32 0
+// CHECK-NEXT: [[VecA1Ptr:%.*]] = getelementptr <3 x float>, ptr [[VecGep1]], i32 0, i32 0
+// CHECK-NEXT: store float [[Val1]], ptr [[VecA1Ptr]], align 4
+
+// CHECK-NEXT: [[VecA2:%.*]] = load <3 x float>, ptr [[VecGep3]], align 4
+// CHECK-NEXT: [[Val2:%.*]] = extractelement <3 x float> [[VecA2]], i32 1
+// CHECK-NEXT: [[VecA2Ptr:%.*]] = getelementptr <3 x float>, ptr [[VecGep1]], i32 0, i32 1
+// CHECK-NEXT: store float [[Val2]], ptr [[VecA2Ptr]], align 4
+
+// CHECK-NEXT: [[VecA3:%.*]] = load <3 x float>, ptr [[VecGep3]], align 4
+// CHECK-NEXT: [[Val3:%.*]] = extractelement <3 x float> [[VecA3]], i32 2
+// CHECK-NEXT: [[VecA3Ptr:%.*]] = getelementptr <3 x float>, ptr [[VecGep1]], i32 0, i32 2
+// CHECK-NEXT: store float [[Val3]], ptr [[VecA3Ptr]], align 4
+
+// CHECK-NEXT: [[VecB1:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
+// CHECK-NEXT: [[Val4:%.*]] = extractelement <4 x float> [[VecB1]], i32 0
+// CHECK-NEXT: [[VecB1Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 0
+// CHECK-NEXT: store float [[Val4]], ptr [[VecB1Ptr]], align 4
+
+// CHECK-NEXT: [[VecB2:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
+// CHECK-NEXT: [[Val5:%.*]] = extractelement <4 x float> [[VecB2]], i32 1
+// CHECK-NEXT: [[VecB2Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 1
+// CHECK-NEXT: store float [[Val5]], ptr [[VecB2Ptr]], align 4
+
+// CHECK-NEXT: [[VecB3:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
+// CHECK-NEXT: [[Val6:%.*]] = extractelement <4 x float> [[VecB3]], i32 2
+// CHECK-NEXT: [[VecB3Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 2
+// CHECK-NEXT: store float [[Val6]], ptr [[VecB3Ptr]], align 4
+
+// CHECK-NEXT: [[VecB4:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
+// CHECK-NEXT: [[Val7:%.*]] = extractelement <4 x float> [[VecB4]], i32 3
+// CHECK-NEXT: [[VecB4Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 3
+// CHECK-NEXT: store float [[Val7]], ptr [[VecB4Ptr]], align 4
+
+// CHECK-NEXT: ret void
+void case2() {
+ S localS = (S)cbs;
}
diff --git a/clang/test/CodeGenHLSL/resources/resources-in-structs.hlsl b/clang/test/CodeGenHLSL/resources/resources-in-structs.hlsl
index 3502f1d7eca66..bf4c2f5558acf 100644
--- a/clang/test/CodeGenHLSL/resources/resources-in-structs.hlsl
+++ b/clang/test/CodeGenHLSL/resources/resources-in-structs.hlsl
@@ -47,7 +47,7 @@ C c : register(t10);
// Check that c.BufOne is initialized from binding with counter
//
-// CHECK: define internal void @__cxx_global_var_init.3()
+// CHECK: define internal void @__cxx_global_var_init.{{[0-9]+}}()
// CHECK-NEXT: entry:
// CHECK-NEXT: call void @hlsl::StructuredBuffer<float>::__createFromBinding(unsigned int, unsigned int, int, unsigned int, char const*)
// CHECK-SAME: (ptr dead_on_unwind writable sret(%"class.hlsl::StructuredBuffer") align 4 @c.BufOne, i32 noundef 16, i32 noundef 0, i32 noundef 1, i32 noundef 0, ptr noundef @[[cBufOne]])
diff --git a/clang/test/CodeGenHLSL/semantics/semantic-struct-2-output.hlsl b/clang/test/CodeGenHLSL/semantics/semantic-struct-2-output.hlsl
index 2f8dc97ef762e..e50044811017a 100644
--- a/clang/test/CodeGenHLSL/semantics/semantic-struct-2-output.hlsl
+++ b/clang/test/CodeGenHLSL/semantics/semantic-struct-2-output.hlsl
@@ -14,8 +14,8 @@ struct Output {
// Make sure SV_DispatchThreadID translated into dx.thread.id.
-// CHECK-DX: define hidden void @_Z3foo5Input(ptr dead_on_unwind noalias writable sret(%struct.Output) align 1 %agg.result, ptr noundef byval(%struct.Input) align 1 %input)
-// CHECK-VK: define hidden spir_func void @_Z3foo5Input(ptr dead_on_unwind noalias writable sret(%struct.Output) align 1 %agg.result, ptr noundef byval(%struct.Input) align 1 %input)
+// CHECK-DX: define hidden void @_Z3foo5Input(ptr dead_on_unwind noalias writable sret(%struct.Output) align 1 %agg.result, ptr noundef dead_on_return %input)
+// CHECK-VK: define hidden spir_func void @_Z3foo5Input(ptr dead_on_unwind noalias writable sret(%struct.Output) align 1 %agg.result, ptr noundef dead_on_return %input)
// CHECK: %Idx = getelementptr inbounds nuw %struct.Input, ptr %input, i32 0, i32 0
// CHECK: %[[#tmp:]] = load float, ptr %Idx, align 1
diff --git a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
index a2df307038774..5cbd7af9b91cd 100644
--- a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
+++ b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
@@ -6,10 +6,10 @@ struct Pair {
int getFirst() {
Pair Another = {5, 10};
this = Another;
- return this.First;
+ return this.First;
}
int getSecond() {
- this = Pair();
+ this = {0, 123};
return Second;
}
void operator=(Pair P) {
@@ -26,30 +26,30 @@ void main() {
// This test makes a probably safe assumption that HLSL 202x includes operator overloading for assignment operators.
// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair8getFirstEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
-// CHECK-NEXT:entry:
-// CHECK-NEXT:%this.addr = alloca ptr, align 4
-// CHECK-NEXT:%Another = alloca %struct.Pair, align 1
-// CHECK-NEXT:%agg.tmp = alloca %struct.Pair, align 1
-// CHECK-NEXT:store ptr %this, ptr %this.addr, align 4
-// CHECK-NEXT:%this1 = load ptr, ptr %this.addr, align 4
-// CHECK-NEXT:%First = getelementptr inbounds nuw %struct.Pair, ptr %Another, i32 0, i32 0
-// CHECK-NEXT:store i32 5, ptr %First, align 1
-// CHECK-NEXT:%Second = getelementptr inbounds nuw %struct.Pair, ptr %Another, i32 0, i32 1
-// CHECK-NEXT:store i32 10, ptr %Second, align 1
-// CHECK-NEXT:call void @llvm.memcpy.p0.p0.i32(ptr align 1 %agg.tmp, ptr align 1 %Another, i32 8, i1 false)
-// CHECK-NEXT:call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) %this1, ptr noundef byval(%struct.Pair) align 1 %agg.tmp)
-// CHECK-NEXT:%First2 = getelementptr inbounds nuw %struct.Pair, ptr %this1, i32 0, i32 0
-// CHECK-NEXT:%0 = load i32, ptr %First2, align 1
-// CHECK-NEXT:ret i32 %0
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[Another:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Another]], ptr align 1 @__const._ZN4Pair8getFirstEv.Another, i32 8, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[Another]], i32 8, i1 false)
+// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
+// CHECK-NEXT: %0 = load i32, ptr [[First]], align 1
+// CHECK-NEXT: ret i32 %0
// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair9getSecondEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
// CHECK-NEXT:entry:
-// CHECK-NEXT:%this.addr = alloca ptr, align 4
-// CHECK-NEXT:%agg.tmp = alloca %struct.Pair, align 1
-// CHECK-NEXT:store ptr %this, ptr %this.addr, align 4
-// CHECK-NEXT:%this1 = load ptr, ptr %this.addr, align 4
-// CHECK-NEXT:call void @llvm.memset.p0.i32(ptr align 1 %agg.tmp, i8 0, i32 8, i1 false)
-// CHECK-NEXT:call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) %this1, ptr noundef byval(%struct.Pair) align 1 %agg.tmp)
-// CHECK-NEXT:%Second = getelementptr inbounds nuw %struct.Pair, ptr %this1, i32 0, i32 1
-// CHECK-NEXT:%0 = load i32, ptr %Second, align 1
-// CHECK-NEXT:ret i32 %0
+// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 0
+// CHECK-NEXT: store i32 0, ptr [[First]], align 1
+// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 1
+// CHECK-NEXT: store i32 123, ptr [[Second]], align 1
+// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
+// CHECK-NEXT: [[Second2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
+// CHECK-NEXT: %0 = load i32, ptr [[Second2]], align 1
+// CHECK-NEXT: ret i32 %0
diff --git a/clang/test/CodeGenHLSL/this-assignment.hlsl b/clang/test/CodeGenHLSL/this-assignment.hlsl
index efccc96499242..8fb0b05737b2e 100644
--- a/clang/test/CodeGenHLSL/this-assignment.hlsl
+++ b/clang/test/CodeGenHLSL/this-assignment.hlsl
@@ -12,7 +12,7 @@ struct Pair {
// In HLSL 202x, this is a move assignment rather than a copy.
int getSecond() {
- this = Pair();
+ this = {0, 123};
return Second;
}
@@ -34,31 +34,42 @@ void main() {
// This tests reference like implicit this in HLSL
// CHECK-LABEL: define {{.*}}getFirst
-// CHECK-NEXT:entry:
-// CHECK-NEXT:%this.addr = alloca ptr, align 4
-// CHECK-NEXT:%Another = alloca %struct.Pair, align 1
-// CHECK-NEXT:store ptr %this, ptr %this.addr, align 4
-// CHECK-NEXT:%this1 = load ptr, ptr %this.addr, align 4
-// CHECK-NEXT:call void @llvm.memcpy.p0.p0.i32(ptr align 1 %Another, ptr align 1 @__const._ZN4Pair8getFirstEv.Another, i32 8, i1 false)
-// CHECK-NEXT:call void @llvm.memcpy.p0.p0.i32(ptr align 1 %this1, ptr align 1 %Another, i32 8, i1 false)
-// CHECK-NEXT:%First = getelementptr inbounds nuw %struct.Pair, ptr %this1, i32 0, i32 0
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[Another:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Another]], ptr align 1 @__const._ZN4Pair8getFirstEv.Another, i32 8, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[ThisPtr]], ptr align 1 [[Another]], i32 8, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[ThisPtr]], i32 8, i1 false)
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
// CHECK-LABEL: define {{.*}}getSecond
-// CHECK-NEXT:entry:
-// CHECK-NEXT:%this.addr = alloca ptr, align 4
-// CHECK-NEXT:%ref.tmp = alloca %struct.Pair, align 1
-// CHECK-NEXT:store ptr %this, ptr %this.addr, align 4
-// CHECK-NEXT:%this1 = load ptr, ptr %this.addr, align 4
-// CHECK-NEXT:call void @llvm.memset.p0.i32(ptr align 1 %ref.tmp, i8 0, i32 8, i1 false)
-// CHECK-NEXT:call void @llvm.memcpy.p0.p0.i32(ptr align 1 %this1, ptr align 1 %ref.tmp, i32 8, i1 false)
-// CHECK-NEXT:%Second = getelementptr inbounds nuw %struct.Pair, ptr %this1, i32 0, i32 1
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[ThisPtrAddr:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAddr]], align 4
+// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAddr]], align 4
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
+// CHECK-NEXT: store i32 0, ptr [[First]], align 1
+// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
+// CHECK-NEXT: store i32 123, ptr [[Second]], align 1
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[ThisPtr]], i32 8, i1 false)
+// CHECK-NEXT: [[Second2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
// CHECK-LABEL: define {{.*}}DoSilly
// CHECK-NEXT:entry:
-// CHECK-NEXT: [[ThisPtrAddr:%.*]] = alloca ptr
+// CHECK-NEXT: [[ResultPtr:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[ThisPtrAddr:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[ObjIndirectAdds:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr {{.*}}, ptr [[ResultPtr]]
// CHECK-NEXT: store ptr {{.*}}, ptr [[ThisPtrAddr]]
+// CHECK-NEXT: store ptr {{.*}}, ptr [[ObjIndirectAdds]]
// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAddr]]
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[ThisPtr]], ptr align 1 [[Obj:%.*]], i32 8, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[ThisPtr]], i32 8, i1 false)
// CHECK-NEXT: [[FirstAddr:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
// CHECK-NEXT: [[First:%.*]] = load i32, ptr [[FirstAddr]]
// CHECK-NEXT: [[FirstPlusTwo:%.*]] = add nsw i32 [[First]], 2
diff --git a/clang/test/SemaHLSL/BuiltIns/WaveActiveAllTrue-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/WaveActiveAllTrue-errors.hlsl
index b0d0fdfca5e18..cb680d01afb0f 100644
--- a/clang/test/SemaHLSL/BuiltIns/WaveActiveAllTrue-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/WaveActiveAllTrue-errors.hlsl
@@ -17,5 +17,5 @@ struct Foo
bool test_type_check(Foo p0) {
return __builtin_hlsl_wave_active_all_true(p0);
- // expected-error at -1 {{no viable conversion from 'Foo' to 'bool'}}
+ // expected-error at -1 {{cannot initialize a parameter of type 'bool' with an lvalue of type 'Foo'}}
}
diff --git a/clang/test/SemaHLSL/BuiltIns/WaveActiveAnyTrue-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/WaveActiveAnyTrue-errors.hlsl
index 875aae0651702..85f0719a8a7a7 100644
--- a/clang/test/SemaHLSL/BuiltIns/WaveActiveAnyTrue-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/WaveActiveAnyTrue-errors.hlsl
@@ -17,5 +17,5 @@ struct Foo
bool test_type_check(Foo p0) {
return __builtin_hlsl_wave_active_any_true(p0);
- // expected-error at -1 {{no viable conversion from 'Foo' to 'bool'}}
+ // expected-error at -1 {{cannot initialize a parameter of type 'bool' with an lvalue of type 'Foo'}}
}
diff --git a/clang/test/SemaHLSL/BuiltIns/WaveActiveBallot-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/WaveActiveBallot-errors.hlsl
index cb9e69bc89c26..802226a935596 100644
--- a/clang/test/SemaHLSL/BuiltIns/WaveActiveBallot-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/WaveActiveBallot-errors.hlsl
@@ -17,5 +17,5 @@ struct Foo
uint4 test_type_check(Foo p0) {
return __builtin_hlsl_wave_active_ballot(p0);
- // expected-error at -1 {{no viable conversion from 'Foo' to 'bool'}}
+ // expected-error at -1 {{cannot initialize a parameter of type 'bool' with an lvalue of type 'Foo'}}
}
diff --git a/clang/test/SemaHLSL/BuiltIns/WaveActiveCountBits-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/WaveActiveCountBits-errors.hlsl
index 02f45eb30b377..da159c5ebd102 100644
--- a/clang/test/SemaHLSL/BuiltIns/WaveActiveCountBits-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/WaveActiveCountBits-errors.hlsl
@@ -14,5 +14,5 @@ struct S { float f; };
int test_bad_conversion(S x) {
return __builtin_hlsl_wave_active_count_bits(x);
- // expected-error at -1 {{no viable conversion from 'S' to 'bool'}}
+ // expected-error at -1 {{cannot initialize a parameter of type 'bool' with an lvalue of type 'S'}}
}
diff --git a/clang/test/SemaHLSL/BuiltIns/dot4add_i8packed-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/dot4add_i8packed-errors.hlsl
index ac0b430bfaf94..58956d1c82cbe 100644
--- a/clang/test/SemaHLSL/BuiltIns/dot4add_i8packed-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/dot4add_i8packed-errors.hlsl
@@ -24,5 +24,5 @@ struct S { float f; };
int test_expr_struct_type_check(S p0, int p1) {
return __builtin_hlsl_dot4add_i8packed(p0, p1, p1);
- // expected-error at -1 {{no viable conversion from 'S' to 'unsigned int'}}
+ // expected-error at -1 {{cannot initialize a parameter of type 'unsigned int' with an lvalue of type 'S'}}
}
diff --git a/clang/test/SemaHLSL/BuiltIns/dot4add_u8packed-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/dot4add_u8packed-errors.hlsl
index f1fa41902b968..cc6441cc88be3 100644
--- a/clang/test/SemaHLSL/BuiltIns/dot4add_u8packed-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/dot4add_u8packed-errors.hlsl
@@ -24,5 +24,5 @@ struct S { float f; };
int test_expr_struct_type_check(S p0, uint p1) {
return __builtin_hlsl_dot4add_u8packed(p1, p1, p0);
- // expected-error at -1 {{no viable conversion from 'S' to 'unsigned int'}}
+ // expected-error at -1 {{cannot initialize a parameter of type 'unsigned int' with an lvalue of type 'S'}}
}
diff --git a/clang/test/SemaHLSL/Language/AggregateSplatCast-errors.hlsl b/clang/test/SemaHLSL/Language/AggregateSplatCast-errors.hlsl
index fbb47bd2e7d39..d180611263fe4 100644
--- a/clang/test/SemaHLSL/Language/AggregateSplatCast-errors.hlsl
+++ b/clang/test/SemaHLSL/Language/AggregateSplatCast-errors.hlsl
@@ -16,7 +16,7 @@ struct R {
// Can't cast a union
export void cantCast2() {
R r = (R)1;
- // expected-error at -1 {{no matching conversion for C-style cast from 'int' to 'R'}}
+ // expected-error at -1 {{C-style cast from 'int' to 'R' is not allowed}}
}
RWBuffer<float4> Buf;
@@ -39,5 +39,5 @@ struct X {
export void cantCast5() {
X x = (X)1;
- // expected-error at -1 {{no matching conversion for C-style cast from 'int' to 'X'}}
+ // expected-error at -1 {{C-style cast from 'int' to 'X' is not allowed}}
}
diff --git a/clang/test/SemaHLSL/Language/ElementwiseCast-errors.hlsl b/clang/test/SemaHLSL/Language/ElementwiseCast-errors.hlsl
index d9f50e9b0307f..b8fb1b4596d2a 100644
--- a/clang/test/SemaHLSL/Language/ElementwiseCast-errors.hlsl
+++ b/clang/test/SemaHLSL/Language/ElementwiseCast-errors.hlsl
@@ -8,9 +8,6 @@ export void cantCast() {
}
struct R {
-// expected-note at -1 {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'int2' (aka 'vector<int, 2>') to 'const R' for 1st argument}}
-// expected-note at -2 {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'int2' (aka 'vector<int, 2>') to 'R' for 1st argument}}
-// expected-note at -3 {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
int A;
union {
float F;
@@ -21,10 +18,10 @@ struct R {
export void cantCast4() {
int2 A = {1,2};
R r = R(A);
- // expected-error at -1 {{no matching conversion for functional-style cast from 'int2' (aka 'vector<int, 2>') to 'R'}}
+ // expected-error at -1 {{functional-style cast from 'int2' (aka 'vector<int, 2>') to 'R' is not allowed}}
R r2;
r2.A = 1;
r2.F = 2.0;
int2 B = (int2)r2;
- // expected-error at -1 {{cannot convert 'R' to 'int2' (aka 'vector<int, 2>') without a conversion operator}}
+ // expected-error at -1 {{C-style cast from 'R' to 'int2' (aka 'vector<int, 2>') is not allowed}}
}
diff --git a/clang/test/SemaHLSL/Language/ElementwiseCasts.hlsl b/clang/test/SemaHLSL/Language/ElementwiseCasts.hlsl
index 8481cfc1b18e2..9d296a9798a9c 100644
--- a/clang/test/SemaHLSL/Language/ElementwiseCasts.hlsl
+++ b/clang/test/SemaHLSL/Language/ElementwiseCasts.hlsl
@@ -30,6 +30,7 @@ struct S {
// cast from a struct
// CHECK-LABEL: call3
// CHECK: CStyleCastExpr {{.*}} 'int[2]' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'S' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue Var {{.*}} 'SS' 'S'
export void call3() {
S SS = {1,1.0};
diff --git a/clang/test/SemaHLSL/Language/InitListAST.hlsl b/clang/test/SemaHLSL/Language/InitListAST.hlsl
index 62acaf3046548..b3325b37a8939 100644
--- a/clang/test/SemaHLSL/Language/InitListAST.hlsl
+++ b/clang/test/SemaHLSL/Language/InitListAST.hlsl
@@ -1076,12 +1076,14 @@ float case17() {
// CHECK-NEXT: OpaqueValueExpr [[OPV0:0x[0-9a-f]+]] {{.*}} 'TwoInts' xvalue
// CHECK-NEXT: MaterializeTemporaryExpr {{.*}} 'TwoInts' xvalue
// CHECK-NEXT: CStyleCastExpr {{.*}} 'TwoInts' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'TwoFloats' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'TwoFloats' lvalue Var {{.*}} 'TF' 'TwoFloats'
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'int' <LValueToRValue>
// CHECK-NEXT: MemberExpr {{.*}} 'int' xvalue .W
// CHECK-NEXT: OpaqueValueExpr [[OPV0]] {{.*}} 'TwoInts' xvalue
// CHECK-NEXT: MaterializeTemporaryExpr {{.*}} 'TwoInts' xvalue
// CHECK-NEXT: CStyleCastExpr {{.*}} 'TwoInts' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'TwoFloats' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'TwoFloats' lvalue Var {{.*}} 'TF' 'TwoFloats'
int case18() {
TwoFloats TF = {1.0,2.0};
@@ -1107,6 +1109,7 @@ int case18() {
// CHECK-NEXT: OpaqueValueExpr [[OPV1:0x[0-9a-f]+]] {{.*}} 'int[4]' xvalue
// CHECK-NEXT: MaterializeTemporaryExpr {{.*}} 'int[4]' xvalue
// CHECK-NEXT: CStyleCastExpr {{.*}} 'int[4]' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'FourFloats' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'FourFloats' lvalue Var {{.*}} 'FF' 'FourFloats'
// CHECK-NEXT: IntegerLiteral {{.*}} '__size_t':'unsigned long' 0
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'int' <LValueToRValue>
@@ -1115,6 +1118,7 @@ int case18() {
// CHECK-NEXT: OpaqueValueExpr [[OPV1]] {{.*}} 'int[4]' xvalue
// CHECK-NEXT: MaterializeTemporaryExpr {{.*}} 'int[4]' xvalue
// CHECK-NEXT: CStyleCastExpr {{.*}} 'int[4]' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'FourFloats' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'FourFloats' lvalue Var {{.*}} 'FF' 'FourFloats'
// CHECK-NEXT: IntegerLiteral {{.*}} '__size_t':'unsigned long' 1
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'int' <LValueToRValue>
@@ -1123,6 +1127,7 @@ int case18() {
// CHECK-NEXT: OpaqueValueExpr [[OPV1]] {{.*}} 'int[4]' xvalue
// CHECK-NEXT: MaterializeTemporaryExpr {{.*}} 'int[4]' xvalue
// CHECK-NEXT: CStyleCastExpr {{.*}} 'int[4]' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'FourFloats' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'FourFloats' lvalue Var {{.*}} 'FF' 'FourFloats'
// CHECK-NEXT: IntegerLiteral {{.*}} '__size_t':'unsigned long' 2
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'int' <LValueToRValue>
@@ -1131,6 +1136,7 @@ int case18() {
// CHECK-NEXT: OpaqueValueExpr [[OPV1]] {{.*}} 'int[4]' xvalue
// CHECK-NEXT: MaterializeTemporaryExpr {{.*}} 'int[4]' xvalue
// CHECK-NEXT: CStyleCastExpr {{.*}} 'int[4]' <HLSLElementwiseCast>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'FourFloats' <LValueToRValue> part_of_explicit_cast
// CHECK-NEXT: DeclRefExpr {{.*}} 'FourFloats' lvalue Var {{.*}} 'FF' 'FourFloats'
// CHECK-NEXT: IntegerLiteral {{.*}} '__size_t':'unsigned long' 3
int case19() {
diff --git a/clang/test/SemaHLSL/Language/InitLists.hlsl b/clang/test/SemaHLSL/Language/InitLists.hlsl
index c31c0fde33f30..1ea9ddb2aff26 100644
--- a/clang/test/SemaHLSL/Language/InitLists.hlsl
+++ b/clang/test/SemaHLSL/Language/InitLists.hlsl
@@ -106,20 +106,14 @@ struct R {
};
};
-// expected-note@#anon{{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'int' to}}
-// expected-note@#anon{{candidate constructor (the implicit move constructor) not viable: no known conversion from 'int' to}}
-
void Err2(RWBuffer<float4> B) {
ContainsResource RS1 = {1, B};
- ContainsResource RS2 = (1.xx); // expected-error{{no viable conversion from 'vector<int, 2>' (vector of 2 'int' values) to 'ContainsResource'}}
+ ContainsResource RS2 = (1.xx); // expected-error{{cannot initialize a variable of type 'ContainsResource' with an rvalue of type 'vector<int, 2>' (vector of 2 'int' values)}}
ContainsResource RS3 = {B, 1}; // expected-error{{no viable conversion from 'RWBuffer<float4>' (aka 'RWBuffer<vector<float, 4>>') to 'int'}}
ContainsResourceInverted IR = {RS1}; // expected-error{{no viable conversion from 'int' to 'hlsl::RWBuffer<vector<float, 4>>'}}
- R r = {1,2}; // expected-error{{no viable conversion from 'int' to 'R::(anonymous union at}}
+ R r = {1,2}; // expected-error-re{{cannot initialize a parameter of type 'R::(anonymous union at {{.+}} with an rvalue of type 'int'}}
}
-// expected-note@#ContainsResource{{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'vector<int, 2>' (vector of 2 'int' values) to 'const ContainsResource &' for 1st argument}}
-// expected-note@#ContainsResource{{candidate constructor (the implicit move constructor) not viable: no known conversion from 'vector<int, 2>' (vector of 2 'int' values) to 'ContainsResource &&' for 1st argument}}
-
// This note refers to the RWBuffer copy constructor that do not have a source locations
// expected-note@*{{candidate constructor not viable}}
diff --git a/clang/test/SemaHLSL/prohibit_pointer.hlsl b/clang/test/SemaHLSL/prohibit_pointer.hlsl
index 76c017150f9d5..84e1e10c45820 100644
--- a/clang/test/SemaHLSL/prohibit_pointer.hlsl
+++ b/clang/test/SemaHLSL/prohibit_pointer.hlsl
@@ -68,7 +68,7 @@ struct Fish {
// expected-note at +1 {{'->' applied to return value of the operator->() declared here}}
Fins operator ->() {
- return Fins();
+ return Fins(); // expected-error {{no matching constructor for initialization of 'Fins'}}
}
};
>From 3691fbaafbfa780563e9c3bea0120fa8452af8e8 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 30 Apr 2026 09:47:21 -0700
Subject: [PATCH 02/22] Update test
---
.../CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
index 8d5ed58e58d74..b76c36de8b418 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
@@ -156,9 +156,12 @@ struct Derived : BFields {
// flatten from a derived struct with bitfields
// CHECK-LABEL: call8
-// CHECK: [[A:%.*]] = alloca [4 x i32], align 4
-// CHECK: [[Tmp:%.*]] = alloca %struct.Derived, align 1
-// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[DIndirectAddr:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[A:%.*]] = alloca [4 x i32], align 4
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
+// CHECK-NEXT: store ptr %D, ptr [[DIndirectAddr]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 0
// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 1
// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 2
>From 000180a3753ebf1caf84a66f10bfb948202cee08 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 30 Apr 2026 10:07:54 -0700
Subject: [PATCH 03/22] Fix ObjCXX test failure
---
clang/lib/Sema/SemaExpr.cpp | 13 ++++++-------
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index cba1f4e9fc6ab..fdabed7472d34 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -16006,9 +16006,6 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
if (!LHSExpr || !RHSExpr)
return ExprError();
- const Type *LHSTy = LHSExpr->getType().getTypePtr();
- const Type *RHSTy = RHSExpr->getType().getTypePtr();
-
// We want to end up calling one of SemaPseudoObject::checkAssignment
// (if the LHS is a pseudo-object), BuildOverloadedBinOp (if
// both expressions are overloadable or either is type-dependent),
@@ -16016,7 +16013,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
// any placeholder types out of the way.
// Handle pseudo-objects in the LHS.
- if (const BuiltinType *pty = LHSTy->getAsPlaceholderType()) {
+ if (const BuiltinType *pty = LHSExpr->getType()->getAsPlaceholderType()) {
// Assignments with a pseudo-object l-value need special analysis.
if (pty->getKind() == BuiltinType::PseudoObject &&
BinaryOperator::isAssignmentOp(Opc))
@@ -16072,7 +16069,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
if (Opc == BO_Assign && pty->getKind() == BuiltinType::Overload) {
if (getLangOpts().CPlusPlus &&
(LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent() ||
- LHSTy->isOverloadableType()))
+ LHSExpr->getType()->isOverloadableType()))
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
return CreateBuiltinBinOp(OpLoc, Opc, LHSExpr, RHSExpr,
@@ -16081,7 +16078,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
// Don't resolve overloads if the other type is overloadable.
if (getLangOpts().CPlusPlus && pty->getKind() == BuiltinType::Overload &&
- LHSTy->isOverloadableType())
+ LHSExpr->getType()->isOverloadableType())
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
ExprResult resolvedRHS = CheckPlaceholderExpr(RHSExpr);
@@ -16090,7 +16087,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
}
if (getLangOpts().HLSL &&
- (LHSTy->isHLSLResourceRecord() || LHSTy->isHLSLResourceRecordArray())) {
+ (LHSExpr->getType()->isHLSLResourceRecord() || LHSExpr->getType()->isHLSLResourceRecordArray())) {
if (!HLSL().CheckResourceBinOp(Opc, LHSExpr, RHSExpr, OpLoc))
return ExprError();
}
@@ -16100,6 +16097,8 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
// or has an overloadable type.
// In HLSL, user-defined structs/classes do not have ctors, dtors or
// overloadable operators.
+ QualType LHSTy = LHSExpr->getType();
+ QualType RHSTy = RHSExpr->getType();
bool IsLHSNonOverloadableHLSLType =
getLangOpts().HLSL && LHSTy->isRecordType() &&
!LHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers();
>From 5aa083e16a31572d07809ff7f546121e735a85ba Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 30 Apr 2026 11:02:29 -0700
Subject: [PATCH 04/22] clang-format
---
clang/lib/Sema/SemaExpr.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index fdabed7472d34..1b796354c1532 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -16086,8 +16086,8 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
RHSExpr = resolvedRHS.get();
}
- if (getLangOpts().HLSL &&
- (LHSExpr->getType()->isHLSLResourceRecord() || LHSExpr->getType()->isHLSLResourceRecordArray())) {
+ if (getLangOpts().HLSL && (LHSExpr->getType()->isHLSLResourceRecord() ||
+ LHSExpr->getType()->isHLSLResourceRecordArray())) {
if (!HLSL().CheckResourceBinOp(Opc, LHSExpr, RHSExpr, OpLoc))
return ExprError();
}
>From a18b5a926bfe07ad1cff5ddfb047801a6c2fbd05 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 30 Apr 2026 17:34:47 -0700
Subject: [PATCH 05/22] cleanup comments
---
clang/lib/CodeGen/CGDecl.cpp | 2 +-
clang/lib/Sema/SemaExpr.cpp | 2 +-
clang/lib/Sema/SemaHLSL.cpp | 4 ++--
clang/lib/Sema/SemaOverload.cpp | 4 ++--
4 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index 28a14072982e8..c6b7ec7ee15c6 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -1531,7 +1531,7 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
((Ty.isPODType(getContext()) || BaseTy->isObjCObjectPointerType() ||
// If HLSL, check if it's a constant initializer anyway because
// POD-ness will no longer be true for user defined structs
- // (since they cannot have constructors or a destructor).
+ // (since they do not have constructors).
(getLangOpts().HLSL && BaseTy->isRecordType())) &&
D.getInit()->isConstantInitializer(getContext())))) {
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 1b796354c1532..5c32697b5195a 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -16095,7 +16095,7 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
if (getLangOpts().CPlusPlus) {
// Otherwise, build an overloaded op if either expression is type-dependent
// or has an overloadable type.
- // In HLSL, user-defined structs/classes do not have ctors, dtors or
+ // In HLSL, user-defined structs/classes do not have ctors or
// overloadable operators.
QualType LHSTy = LHSExpr->getType();
QualType RHSTy = RHSExpr->getType();
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 688bb2e39cea8..2a45ce415a435 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -5563,9 +5563,9 @@ bool SemaHLSL::ActOnUninitializedVarDecl(VarDecl *VD) {
}
// User-defined structs/classes do not have constructors.
- // When declared at the a scope, they are part of the constant buffer
+ // When declared at a global scope, they are part of the constant buffer
// and should not be initialized by the compiler.
- // When declated at a local scope, they are by default not initialized.
+ // When declared at a local scope, they are not initialized.
// Also applies to arrays of user-defined structs/classes.
const Type *Ty = VD->getType()->getUnqualifiedDesugaredType();
while (Ty->isArrayType())
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index f93313392e3e7..cdbcc70fa74d7 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -15523,8 +15523,8 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
// various built-in candidates, but as DR507 points out, this can lead to
// problems. So we do it this way, which pretty much follows what GCC does.
// Note that we go the traditional code path for compound assignment forms.
- // In HLSL, user-defined structs/classes do not have ctors, dtors or
- // overloadable operators, so we can take this shortcut too.
+ // In HLSL, user-defined structs/classes do not have or overloadable
+ // operators, so we can take this shortcut too.
const Type *LHSTy = Args[0]->getType().getTypePtr();
if (Opc == BO_Assign &&
(!LHSTy->isOverloadableType() ||
>From cacebdde4879156a4a2810f987cb469a6f03ae75 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 1 May 2026 12:21:20 -0700
Subject: [PATCH 06/22] Add test for cbuffer struct with array and embedded
struct
---
.../resources/cbuffer_struct_passing.hlsl | 200 +++++++++++++-----
1 file changed, 144 insertions(+), 56 deletions(-)
diff --git a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
index 2f7dc02c6d7f2..7858fbcf68543 100644
--- a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
+++ b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
@@ -9,15 +9,25 @@ struct P {
};
struct S : P {
- float4 b;
+ double b;
+ float4 c;
};
-// CHECK-DAG: %S = type <{ <3 x float>, target("[[PADDING_TYPE]]", 4), <4 x float> }>
+struct T {
+ S s;
+ int arr[2];
+};
+
+// CHECK-DAG: %__cblayout_CB = type <{ %S, %T }>
+// CHECK-DAG: %S = type <{ <3 x float>, target("[[PADDING_TYPE]]", 4), double, target("[[PADDING_TYPE]]", 8), <4 x float> }>
+// CHECK-DAG: %T = type <{ %S, <{ [1 x <{ i32, target("[[PADDING_TYPE]]", 12) }>], i32 }> }>
+// CHECK-DAG: %struct.S = type <{ %struct.P, double, <4 x float> }>
// CHECK-DAG: %struct.P = type { <3 x float> }
-// CHECK-DAG: %struct.S = type { %struct.P, <4 x float> }
+// CHECK-DAG: %struct.T = type { %struct.S, [2 x i32] }
cbuffer CB {
S cbs;
+ T cbt;
};
// CHECK-DAG: @cbs = external hidden addrspace([[CONST_ADDR_SPACE]]) global %S, align 1
@@ -27,13 +37,19 @@ cbuffer CB {
//
// Copy S field by field into local variable in default address space.
//
-// CHECK-NEXT: [[LocalS:%.*]] = alloca %struct.S, align 1
-// CHECK-NEXT: [[PtrA:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 0
-// CHECK-NEXT: [[CBufLoad1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
-// CHECK-NEXT: store <3 x float> [[CBufLoad1]], ptr [[PtrA]], align 4
-// CHECK-NEXT: [[PtrB:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 1
-// CHECK-NEXT: [[CBufLoad2:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 4
-// CHECK-NEXT: store <4 x float> [[CBufLoad2]], ptr [[PtrB]], align 4
+// CHECK-NEXT: [[AggTemp:%.*]] = alloca %struct.S, align 1
+
+// CHECK-NEXT: [[Ptr_a:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
+// CHECK-NEXT: [[CBufLoad_a:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
+// CHECK-NEXT: store <3 x float> [[CBufLoad_a]], ptr [[Ptr_a]], align 4
+
+// CHECK-NEXT: [[Ptr_b:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+// CHECK-NEXT: [[CBufLoad_b:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 8
+// CHECK-NEXT: store double [[CBufLoad_b]], ptr [[Ptr_b]], align 8
+
+// CHECK-NEXT: [[Ptr_c:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 2
+// CHECK-NEXT: [[CBufLoad_c:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 32), align 4
+// CHECK-NEXT: store <4 x float> [[CBufLoad_c]], ptr [[Ptr_c]], align 4
// CHECK-NEXT: ret void
void case1() {
S local = cbs;
@@ -47,56 +63,128 @@ void case1() {
//
// CHECK-NEXT: [[LocalS:%.*]] = alloca %struct.S, align 1
// CHECK-NEXT: [[AggTemp:%.*]] = alloca %struct.S, align 1
-// CHECK-NEXT: [[PtrA:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
-// CHECK-NEXT: [[CBufLoad1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
-// CHECK-NEXT: store <3 x float> [[CBufLoad1]], ptr [[PtrA]], align 4
-// CHECK-NEXT: [[PtrB:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
-// CHECK-NEXT: [[CBufLoad2:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 4
-// CHECK-NEXT: store <4 x float> [[CBufLoad2]], ptr [[PtrB]], align 4
+
+// CHECK-NEXT: [[Ptr_a:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
+// CHECK-NEXT: [[CBufLoad_a:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
+// CHECK-NEXT: store <3 x float> [[CBufLoad_a]], ptr [[Ptr_a]], align 4
+
+// CHECK-NEXT: [[Ptr_b:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+// CHECK-NEXT: [[CBufLoad_b:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 8
+// CHECK-NEXT: store double [[CBufLoad_b]], ptr [[Ptr_b]], align 8
+
+// CHECK-NEXT: [[Ptr_c:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 2
+// CHECK-NEXT: [[CBufLoad_c:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 32), align 4
+// CHECK-NEXT: store <4 x float> [[CBufLoad_c]], ptr [[Ptr_c]], align 4
//
// The proces HLSLElementwiseCast - copy individual vector elements between the structs.
//
-// CHECK-NEXT: [[VecGep1:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 0
-// CHECK-NEXT: [[VecGep2:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 1
-// CHECK-NEXT: [[VecGep3:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
-// CHECK-NEXT: [[VecGep4:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
-
-// CHECK-NEXT: [[VecA1:%.*]] = load <3 x float>, ptr [[VecGep3]], align 4
-// CHECK-NEXT: [[Val1:%.*]] = extractelement <3 x float> [[VecA1]], i32 0
-// CHECK-NEXT: [[VecA1Ptr:%.*]] = getelementptr <3 x float>, ptr [[VecGep1]], i32 0, i32 0
-// CHECK-NEXT: store float [[Val1]], ptr [[VecA1Ptr]], align 4
-
-// CHECK-NEXT: [[VecA2:%.*]] = load <3 x float>, ptr [[VecGep3]], align 4
-// CHECK-NEXT: [[Val2:%.*]] = extractelement <3 x float> [[VecA2]], i32 1
-// CHECK-NEXT: [[VecA2Ptr:%.*]] = getelementptr <3 x float>, ptr [[VecGep1]], i32 0, i32 1
-// CHECK-NEXT: store float [[Val2]], ptr [[VecA2Ptr]], align 4
-
-// CHECK-NEXT: [[VecA3:%.*]] = load <3 x float>, ptr [[VecGep3]], align 4
-// CHECK-NEXT: [[Val3:%.*]] = extractelement <3 x float> [[VecA3]], i32 2
-// CHECK-NEXT: [[VecA3Ptr:%.*]] = getelementptr <3 x float>, ptr [[VecGep1]], i32 0, i32 2
-// CHECK-NEXT: store float [[Val3]], ptr [[VecA3Ptr]], align 4
-
-// CHECK-NEXT: [[VecB1:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
-// CHECK-NEXT: [[Val4:%.*]] = extractelement <4 x float> [[VecB1]], i32 0
-// CHECK-NEXT: [[VecB1Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 0
-// CHECK-NEXT: store float [[Val4]], ptr [[VecB1Ptr]], align 4
-
-// CHECK-NEXT: [[VecB2:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
-// CHECK-NEXT: [[Val5:%.*]] = extractelement <4 x float> [[VecB2]], i32 1
-// CHECK-NEXT: [[VecB2Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 1
-// CHECK-NEXT: store float [[Val5]], ptr [[VecB2Ptr]], align 4
-
-// CHECK-NEXT: [[VecB3:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
-// CHECK-NEXT: [[Val6:%.*]] = extractelement <4 x float> [[VecB3]], i32 2
-// CHECK-NEXT: [[VecB3Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 2
-// CHECK-NEXT: store float [[Val6]], ptr [[VecB3Ptr]], align 4
-
-// CHECK-NEXT: [[VecB4:%.*]] = load <4 x float>, ptr [[VecGep4]], align 4
-// CHECK-NEXT: [[Val7:%.*]] = extractelement <4 x float> [[VecB4]], i32 3
-// CHECK-NEXT: [[VecB4Ptr:%.*]] = getelementptr <4 x float>, ptr [[VecGep2]], i32 0, i32 3
-// CHECK-NEXT: store float [[Val7]], ptr [[VecB4Ptr]], align 4
+// CHECK-NEXT: [[Ptr_LocalS_a:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 0, i32 0
+// CHECK-NEXT: [[Ptr_LocalS_b:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 1
+// CHECK-NEXT: [[Ptr_LocalS_c:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 2
+// CHECK-NEXT: [[Ptr_AggTemp_a:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0, i32 0
+// CHECK-NEXT: [[Ptr_AggTemp_b:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+// CHECK-NEXT: [[Ptr_AggTemp_c:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 2
+
+// CHECK-NEXT: [[AggTemp_a:%.*]] = load <3 x float>, ptr [[Ptr_AggTemp_a]], align 4
+// CHECK-NEXT: [[Val_a0:%.*]] = extractelement <3 x float> [[AggTemp_a]], i32 0
+// CHECK-NEXT: [[Ptr_LocalS_a0:%.*]] = getelementptr <3 x float>, ptr [[Ptr_LocalS_a]], i32 0, i32 0
+// CHECK-NEXT: store float [[Val_a0]], ptr [[Ptr_LocalS_a0]], align 4
+
+// CHECK-NEXT: [[AggTemp_a:%.*]] = load <3 x float>, ptr [[Ptr_AggTemp_a]], align 4
+// CHECK-NEXT: [[Val_a1:%.*]] = extractelement <3 x float> [[AggTemp_a]], i32 1
+// CHECK-NEXT: [[Ptr_LocalS_a1:%.*]] = getelementptr <3 x float>, ptr [[Ptr_LocalS_a]], i32 0, i32 1
+// CHECK-NEXT: store float [[Val_a1]], ptr [[Ptr_LocalS_a1]], align 4
+
+// CHECK-NEXT: [[AggTemp_a:%.*]] = load <3 x float>, ptr [[Ptr_AggTemp_a]], align 4
+// CHECK-NEXT: [[Val_a2:%.*]] = extractelement <3 x float> [[AggTemp_a]], i32 2
+// CHECK-NEXT: [[Ptr_LocalS_a2:%.*]] = getelementptr <3 x float>, ptr [[Ptr_LocalS_a]], i32 0, i32 2
+// CHECK-NEXT: store float [[Val_a2]], ptr [[Ptr_LocalS_a2]], align 4
+
+// CHECK-NEXT: [[Val_b:%.*]] = load double, ptr [[Ptr_AggTemp_b]], align 8
+// CHECK-NEXT: store double [[Val_b]], ptr [[Ptr_LocalS_b]], align 8
+
+// CHECK-NEXT: [[AggTemp_c:%.*]] = load <4 x float>, ptr [[Ptr_AggTemp_c]], align 4
+// CHECK-NEXT: [[Val_c0:%.*]] = extractelement <4 x float> [[AggTemp_c]], i32 0
+// CHECK-NEXT: [[Ptr_LocalS_c0:%.*]] = getelementptr <4 x float>, ptr [[Ptr_LocalS_c]], i32 0, i32 0
+// CHECK-NEXT: store float [[Val_c0]], ptr [[Ptr_LocalS_c0]], align 4
+
+// CHECK-NEXT: [[AggTemp_c:%.*]] = load <4 x float>, ptr [[Ptr_AggTemp_c]], align 4
+// CHECK-NEXT: [[Val_c1:%.*]] = extractelement <4 x float> [[AggTemp_c]], i32 1
+// CHECK-NEXT: [[Ptr_LocalS_c1:%.*]] = getelementptr <4 x float>, ptr [[Ptr_LocalS_c]], i32 0, i32 1
+// CHECK-NEXT: store float [[Val_c1]], ptr [[Ptr_LocalS_c1]], align 4
+
+// CHECK-NEXT: [[AggTemp_c:%.*]] = load <4 x float>, ptr [[Ptr_AggTemp_c]], align 4
+// CHECK-NEXT: [[Val_c2:%.*]] = extractelement <4 x float> [[AggTemp_c]], i32 2
+// CHECK-NEXT: [[Ptr_LocalS_c2:%.*]] = getelementptr <4 x float>, ptr [[Ptr_LocalS_c]], i32 0, i32 2
+// CHECK-NEXT: store float [[Val_c2]], ptr [[Ptr_LocalS_c2]], align 4
+
+// CHECK-NEXT: [[AggTemp_c:%.*]] = load <4 x float>, ptr [[Ptr_AggTemp_c]], align 4
+// CHECK-NEXT: [[Val_c3:%.*]] = extractelement <4 x float> [[AggTemp_c]], i32 3
+// CHECK-NEXT: [[Ptr_LocalS_c3:%.*]] = getelementptr <4 x float>, ptr [[Ptr_LocalS_c]], i32 0, i32 3
+// CHECK-NEXT: store float [[Val_c3]], ptr [[Ptr_LocalS_c3]], align 4
// CHECK-NEXT: ret void
void case2() {
- S localS = (S)cbs;
+ S AggTemp = (S)cbs;
+}
+
+// CHECK-LABEL: case3
+// CHECK-NEXT: entry:
+// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+void case3() {
+
+// CHECK-NEXT: [[LocalT:%.*]] = alloca %struct.T, align 1
+// CHECK-NEXT: [[LocalTCopy:%.*]] = alloca %struct.T, align 1
+// CHECK-NEXT: [[AggTemp:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[AggTempCopy:%.*]] = alloca %struct.S, align 1
+
+// Check that constant to default address space copies the struct field by field
+//
+// CHECK-NEXT: [[Ptr_s:%.*]] = getelementptr inbounds %struct.T, ptr [[LocalT]], i32 0, i32 0
+// CHECK-NEXT: [[Ptr_a:%.*]] = getelementptr inbounds %struct.S, ptr [[Ptr_s]], i32 0, i32 0
+// CHECK-NEXT: [[CbufLoad_a:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, align 4
+// CHECK-NEXT: store <3 x float> [[CbufLoad_a]], ptr [[Ptr_a]], align 4
+
+// CHECK-NEXT: [[Ptr_b:%.*]] = getelementptr inbounds %struct.S, ptr [[Ptr_s]], i32 0, i32 1
+// CHECK-NEXT: [[CbufLoad_c:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 16), align 8
+// CHECK-NEXT: store double [[CbufLoad_c]], ptr [[Ptr_b]], align 8
+
+// CHECK-NEXT: [[Ptr_c:%.*]] = getelementptr inbounds %struct.S, ptr [[Ptr_s]], i32 0, i32 2
+// CHECK-NEXT: [[CbufLoad_b:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 32), align 4
+// CHECK-NEXT: store <4 x float> [[CbufLoad_b]], ptr [[Ptr_c]], align 4
+
+// CHECK-NEXT: [[Ptr_arr:%.*]] = getelementptr inbounds %struct.T, ptr [[LocalT]], i32 0, i32 1
+// CHECK-NEXT: [[Ptr_arr0:%.*]] = getelementptr inbounds [2 x i32], ptr [[Ptr_arr]], i32 0, i32 0
+// CHECK-NEXT: [[CbufLoad_arr0:%.*]] = load i32, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 48), align 4
+// CHECK-NEXT: store i32 [[CbufLoad_arr0]], ptr [[Ptr_arr0]], align 4
+
+// CHECK-NEXT: [[Ptr_arr1:%.*]] = getelementptr inbounds [2 x i32], ptr [[Ptr_arr]], i32 0, i32 1
+// CHECK-NEXT: [[CbufLoad_arr1:%.*]] = load i32, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 64), align 4
+// CHECK-NEXT: store i32 [[CbufLoad_arr1]], ptr [[Ptr_arr1]], align 4
+ T localT = cbt;
+
+// Check that default to default address space copy uses memcpy
+//
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[LocalTCopy]], ptr align 1 [[LocalT]], {{i32|i64}} 44, i1 false)
+ T localTCopy = localT;
+
+// Check that constant to default address space copies the struct field by field
+//
+// CHECK-NEXT: [[Ptr_a1:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
+// CHECK-NEXT: [[CbufLoad_a1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, align 4
+// CHECK-NEXT: store <3 x float> [[CbufLoad_a1]], ptr [[Ptr_a1]], align 4
+// CHECK-NEXT: [[Ptr_b1:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+// CHECK-NEXT: [[CbufLoad_b1:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 16), align 8
+// CHECK-NEXT: store double [[CbufLoad_b1]], ptr [[Ptr_b1]], align 8
+// CHECK-NEXT: [[Ptr_c1:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 2
+// CHECK-NEXT: [[CbufLoad_c1:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 32), align 4
+// CHECK-NEXT: store <4 x float> [[CbufLoad_c1]], ptr [[Ptr_c1]], align 4
+ S AggTemp = cbt.s;
+
+// Check that default to default address space copy uses memcpy
+//
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[AggTempCopy]], ptr align 1 [[AggTemp]], {{i32|i64}} 36, i1 false)
+ S AggTempCopy = AggTemp;
+
+// CHECK-NEXT: ret void
}
>From b8ff4c1453c1c8021f0562d79880c35094660def Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Wed, 6 May 2026 10:56:56 -0700
Subject: [PATCH 07/22] code review feedback
---
clang/include/clang/AST/DeclCXX.h | 81 +++++++++++++++--------------
clang/lib/CodeGen/CGHLSLRuntime.cpp | 12 +++--
clang/lib/Sema/SemaOverload.cpp | 4 +-
3 files changed, 52 insertions(+), 45 deletions(-)
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index fe2bc0dd628c3..2a73e56f2465f 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -768,9 +768,7 @@ class CXXRecordDecl : public RecordDecl {
//
/// Determines whether this class has any user provided special members.
bool hasUserProvidedSpecialMembers() const {
- return data().UserDeclaredSpecialMembers &
- (SMF_MoveConstructor | SMF_MoveAssignment | SMF_Destructor |
- SMF_CopyAssignment | SMF_CopyConstructor) ||
+ return data().UserDeclaredSpecialMembers & SMF_All ||
data().UserDeclaredConstructor ||
data().UserProvidedDefaultConstructor;
}
@@ -780,19 +778,19 @@ class CXXRecordDecl : public RecordDecl {
///
/// This value is used for lazy creation of default constructors.
bool needsImplicitDefaultConstructor() const {
- return ((!data().UserDeclaredConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
- // FIXME: Proposed fix to core wording issue: if a class inherits
- // a default constructor and doesn't explicitly declare one, one
- // is declared implicitly.
- (data().HasInheritedDefaultConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor))) &&
- // In HLSL, only built-in records like resources classes can have
- // constructors.
- (!getLangOpts().HLSL ||
- (isLambda() && lambdaIsDefaultConstructibleAndAssignable()) ||
- hasUserProvidedSpecialMembers());
+ // In HLSL, only built-in records like resources classes can have
+ // constructors and overloadable operators.
+ if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
+ return false;
+
+ return (!data().UserDeclaredConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
+ // FIXME: Proposed fix to core wording issue: if a class inherits
+ // a default constructor and doesn't explicitly declare one, one
+ // is declared implicitly.
+ (data().HasInheritedDefaultConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor));
}
/// Determine whether this class has any user-declared constructors.
@@ -818,11 +816,12 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// constructor to be lazily declared.
bool needsImplicitCopyConstructor() const {
- return !(data().DeclaredSpecialMembers & SMF_CopyConstructor) &&
- // In HLSL, only built-in records like resources classes can have
- // constructors.
- (!getLangOpts().HLSL || isLambda() ||
- hasUserProvidedSpecialMembers());
+ // In HLSL, only built-in records like resources classes can have
+ // constructors and overloadable operators.
+ if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
+ return false;
+
+ return !(data().DeclaredSpecialMembers & SMF_CopyConstructor);
}
/// Determine whether we need to eagerly declare a defaulted copy
@@ -915,14 +914,16 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class should get an implicit move
/// constructor or if any existing special member function inhibits this.
bool needsImplicitMoveConstructor() const {
+ // In HLSL, only built-in records like resources classes can have
+ // constructors and overloadable operators.
+ if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
+ return false;
+
return !(data().DeclaredSpecialMembers & SMF_MoveConstructor) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveAssignment() && !hasUserDeclaredDestructor() &&
- // In HLSL, only built-in records like resources classes can have
- // constructors.
- (!getLangOpts().HLSL || isLambda() ||
- hasUserProvidedSpecialMembers());
+ !hasUserDeclaredMoveAssignment() &&
+ !hasUserDeclaredDestructor();
}
/// Determine whether we need to eagerly declare a defaulted move
@@ -951,11 +952,12 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// assignment operator to be lazily declared.
bool needsImplicitCopyAssignment() const {
- return !(data().DeclaredSpecialMembers & SMF_CopyAssignment) &&
- // In HLSL, only built-in records like resources classes can have
- // constructors.
- (!getLangOpts().HLSL || isLambda() ||
- hasUserProvidedSpecialMembers());
+ // In HLSL, only built-in records like resources classes can have
+ // constructors and overloadable operators.
+ if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
+ return false;
+
+ return !(data().DeclaredSpecialMembers & SMF_CopyAssignment);
}
/// Determine whether we need to eagerly declare a defaulted copy
@@ -1013,17 +1015,18 @@ class CXXRecordDecl : public RecordDecl {
/// assignment operator or if any existing special member function inhibits
/// this.
bool needsImplicitMoveAssignment() const {
+ // In HLSL, only built-in records like resources classes can have
+ // constructors and overloadable operators.
+ if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
+ return false;
+
return !(data().DeclaredSpecialMembers & SMF_MoveAssignment) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveConstructor() && !hasUserDeclaredDestructor() &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable()) &&
- // In HLSL, only built-in records like resources classes can have
- // constructors.
- (!getLangOpts().HLSL ||
- (isLambda() && lambdaIsDefaultConstructibleAndAssignable()) ||
- hasUserProvidedSpecialMembers());
- }
+ !hasUserDeclaredMoveConstructor() &&
+ !hasUserDeclaredDestructor() &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable());
+ }
/// Determine whether we need to eagerly declare a move assignment
/// operator for this class.
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.cpp b/clang/lib/CodeGen/CGHLSLRuntime.cpp
index 575592dc61fec..538dda8ed0881 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.cpp
+++ b/clang/lib/CodeGen/CGHLSLRuntime.cpp
@@ -1079,10 +1079,14 @@ void CGHLSLRuntime::emitEntryFunction(const FunctionDecl *FD,
PD->getAttr<HLSLParamModifierAttr>()) {
llvm_unreachable("Not handled yet");
} else {
- llvm::Type *ParamType = Param.hasByValAttr() ? Param.getParamByValType()
- : PD->getType()->isRecordType()
- ? CGM.getTypes().ConvertType(PD->getType())
- : Param.getType();
+ llvm::Type *ParamType = nullptr;
+ if (Param.hasByValAttr())
+ ParamType = Param.getParamByValType();
+ else if (PD->getType()->isRecordType())
+ ParamType = CGM.getTypes().ConvertType(PD->getType());
+ else
+ ParamType = Param.getType();
+
auto AttrBegin = PD->specific_attr_begin<HLSLAppliedSemanticAttr>();
auto AttrEnd = PD->specific_attr_end<HLSLAppliedSemanticAttr>();
auto Result =
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index cdbcc70fa74d7..5f19e7ae803e5 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -15523,8 +15523,8 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
// various built-in candidates, but as DR507 points out, this can lead to
// problems. So we do it this way, which pretty much follows what GCC does.
// Note that we go the traditional code path for compound assignment forms.
- // In HLSL, user-defined structs/classes do not have or overloadable
- // operators, so we can take this shortcut too.
+ // In HLSL, user-defined structs/classes do not have constructors or
+ // overloadable operators, so we can take this shortcut too.
const Type *LHSTy = Args[0]->getType().getTypePtr();
if (Opc == BO_Assign &&
(!LHSTy->isOverloadableType() ||
>From e46f0e9abcd6588eaadaa1f585a6450b357ba7c7 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 15 May 2026 13:24:50 -0700
Subject: [PATCH 08/22] Handle assignment from derived to base class. Add test
for this, plus AST struct tests.
---
clang/lib/CodeGen/CGExprAgg.cpp | 35 ++++-
clang/lib/Sema/SemaExpr.cpp | 14 ++
clang/test/AST/HLSL/StructPassing-AST.hlsl | 128 ++++++++++++++++++
.../BasicFeatures/StructPassing.hlsl | 113 ++++++++++++++++
.../resources/cbuffer_struct_passing.hlsl | 67 +++++++--
5 files changed, 347 insertions(+), 10 deletions(-)
create mode 100644 clang/test/AST/HLSL/StructPassing-AST.hlsl
create mode 100644 clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl
diff --git a/clang/lib/CodeGen/CGExprAgg.cpp b/clang/lib/CodeGen/CGExprAgg.cpp
index a4282c4f51199..b4e9f9a567621 100644
--- a/clang/lib/CodeGen/CGExprAgg.cpp
+++ b/clang/lib/CodeGen/CGExprAgg.cpp
@@ -868,7 +868,40 @@ void AggExprEmitter::VisitCastExpr(CastExpr *E) {
break;
}
- case CK_DerivedToBase:
+ case CK_DerivedToBase: {
+ assert(CGF.getLangOpts().HLSL &&
+ "Derived/Base casts in EmitAggExpr are only supported in HLSL");
+
+ // Create a temporary for the derived record, switch it out with the current
+ // Dest slot, and emit the derived value.
+ QualType DerivedTy = E->getSubExpr()->getType();
+ AggValueSlot DerivedTmpSlot = CGF.CreateAggTemp(DerivedTy, "tmp");
+
+ AggValueSlot DestBaseSlot = Dest;
+ Dest = DerivedTmpSlot;
+
+ Visit(E->getSubExpr());
+
+ // Perform derived-to-base address conversion to get the address
+ // of the base record within the derived record. In HLSL this should
+ // always be same as the derived because of single inheritance, but let's
+ // do it properly.
+ Address BaseAddrInDerived = CGF.GetAddressOfBaseClass(
+ DerivedTmpSlot.getAddress(), DerivedTy->castAsCXXRecordDecl(),
+ E->path_begin(), E->path_end(),
+ /*NullCheckValue=*/false, E->getExprLoc());
+
+ AggValueSlot SrcBaseSlot = AggValueSlot::forAddr(
+ BaseAddrInDerived, E->getType().getQualifiers(),
+ AggValueSlot::IsNotDestructed, AggValueSlot::DoesNotNeedGCBarriers,
+ AggValueSlot::IsNotAliased, AggValueSlot::DoesNotOverlap);
+
+ // Copy the base class to the original destination slot and restore it.
+ EmitCopy(E->getType(), DestBaseSlot, SrcBaseSlot);
+ Dest = DestBaseSlot;
+ break;
+ }
+
case CK_BaseToDerived:
case CK_UncheckedDerivedToBase: {
llvm_unreachable("cannot perform hierarchy conversion in EmitAggExpr: "
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 5c32697b5195a..b72406f169909 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -10193,6 +10193,20 @@ AssignConvertType Sema::CheckSingleAssignmentConstraints(QualType LHSType,
return AssignConvertType::Incompatible;
}
+ // For HLSL records, insert derived-to-base conversion if needed.
+ if (getLangOpts().HLSL && LHSType->isRecordType()) {
+ QualType RHSType = RHS.get()->getType();
+ if (!Context.hasSameUnqualifiedType(RHSType, LHSType)) {
+ CXXBasePaths Paths;
+ if (IsDerivedFrom(RHS.get()->getBeginLoc(), RHSType, LHSType, Paths)) {
+ CXXCastPath CastPath;
+ BuildBasePathArray(Paths, CastPath);
+ RHS = ImpCastExprToType(RHS.get(), LHSType, CK_DerivedToBase, VK_LValue,
+ &CastPath);
+ }
+ }
+ }
+
// This check seems unnatural, however it is necessary to ensure the proper
// conversion of functions/arrays. If the conversion were done for all
// DeclExpr's (created by ActOnIdExpression), it would mess up the unary
diff --git a/clang/test/AST/HLSL/StructPassing-AST.hlsl b/clang/test/AST/HLSL/StructPassing-AST.hlsl
new file mode 100644
index 0000000000000..8b6b4bbe9a824
--- /dev/null
+++ b/clang/test/AST/HLSL/StructPassing-AST.hlsl
@@ -0,0 +1,128 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -ast-dump -o - %s | FileCheck %s
+
+// CHECK: CXXRecordDecl {{.*}} struct P definition
+// CHECK-NEXT: DefinitionData aggregate standard_layout trivially_copyable pod literal can_const_default_init
+// CHECK-NEXT: DefaultConstructor
+// CHECK-NEXT: CopyConstructor simple trivial implicit_has_const_param
+// CHECK-NEXT: MoveConstructor
+// CHECK-NEXT: CopyAssignment simple trivial implicit_has_const_param
+// CHECK-NEXT: MoveAssignment
+// CHECK-NEXT: Destructor simple irrelevant trivial needs_implicit
+// CHECK-NEXT: PackedAttr
+// CHECK-NEXT: CXXRecordDecl {{.*}} struct P
+// CHECK-NEXT: FieldDecl {{.*}} a 'float'
+// CHECK-NOT: CXXConstructorDecl
+// CHECK-NOT: CXXMethodDecl {{.*}} operator=
+struct P {
+ float a;
+};
+
+// CHECK: CXXRecordDecl {{.*}} struct S definition
+// CHECK-NEXT: DefinitionData aggregate trivially_copyable literal can_const_default_init
+// CHECK-NEXT: DefaultConstructor
+// CHECK-NEXT: CopyConstructor simple trivial
+// CHECK-NEXT: MoveConstructor needs_overload_resolution
+// CHECK-NEXT: CopyAssignment simple trivial
+// CHECK-NEXT: MoveAssignment needs_overload_resolution
+// CHECK-NEXT: Destructor simple irrelevant trivial needs_implicit
+// CHECK-NEXT: public 'P'
+// CHECK-NEXT: PackedAttr
+// CHECK-NEXT: CXXRecordDecl {{.*}} implicit struct S
+// CHECK-NEXT: FieldDecl {{.*}} b 'double'
+// CHECK-NEXT: FieldDecl {{.*}} c 'int[2]'
+// CHECK-NOT: CXXConstructorDecl
+// CHECK-NOT: CXXMethodDecl {{.*}} operator=
+struct S : P {
+ double b;
+ int c[2];
+};
+
+// CHECK: FunctionDecl {{.*}} case1 'void (S)'
+// CHECK-NEXT: ParmVarDecl {{.*}} s 'S'
+// CHECK-NEXT: CompoundStmt
+// CHECK-NEXT: DeclStmt
+// CHECK-NEXT: VarDecl {{.*}} sLocal 'S'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'S' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue ParmVar {{.*}} 's' 'S'
+void case1(S s) {
+ // struct initialization
+ S sLocal = s;
+}
+
+// CHECK: FunctionDecl {{.*}} case2 'void (S)'
+// CHECK-NEXT: ParmVarDecl {{.*}} s 'S'
+// CHECK-NEXT: CompoundStmt
+// CHECK-NEXT: DeclStmt
+// CHECK-NEXT: VarDecl {{.*}} sLocal 'S'
+// CHECK-NEXT: BinaryOperator {{.*}} 'S' lvalue '='
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue Var {{.*}} 'sLocal' 'S'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'S' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue ParmVar {{.*}} 's' 'S'
+void case2(S s) {
+ S sLocal;
+ // struct assignment
+ sLocal = s;
+}
+
+void useS(S s) {}
+
+// CHECK: FunctionDecl {{.*}} case3 'void (S)'
+// CHECK-NEXT: ParmVarDecl {{.*}} used s 'S'
+// CHECK-NEXT: CompoundStmt
+// CHECK-NEXT: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(S)' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (S)' lvalue Function {{.*}} 'useS' 'void (S)'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'S' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue ParmVar {{.*}} 's' 'S'
+void case3(S s) {
+ // struct argument passing
+ useS(s);
+}
+
+// CHECK: FunctionDecl {{.*}} case4 'void (S)'
+// CHECK-NEXT: ParmVarDecl {{.*}} used s 'S'
+// CHECK-NEXT: CompoundStmt
+// CHECK-NEXT: DeclStmt
+// CHECK-NEXT: VarDecl {{.*}} pLocal 'P'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'P' <DerivedToBase (P)>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'S' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue ParmVar {{.*}} 's' 'S'
+void case4(S s) {
+ // derived to base conversion in initialization
+ P pLocal = s;
+}
+
+// CHECK: FunctionDecl {{.*}} case5 'void (S)'
+// CHECK-NEXT: ParmVarDecl {{.*}} used s 'S'
+// CHECK-NEXT: CompoundStmt
+// CHECK-NEXT: DeclStmt
+// CHECK-NEXT: VarDecl {{.*}} pLocal 'P'
+// CHECK-NEXT: BinaryOperator {{.*}} 'P' lvalue '='
+// CHECK-NEXT: DeclRefExpr {{.*}} 'P' lvalue Var {{.*}} 'pLocal' 'P'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'P' <LValueToRValue>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'P' lvalue <DerivedToBase (P)>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue ParmVar {{.*}} 's' 'S'
+void case5(S s) {
+ P pLocal;
+ // derived to base conversion in assignment
+ pLocal = s;
+}
+
+void useP(P p) {}
+
+// CHECK: FunctionDecl {{.*}} case6 'void (S)'
+// CHECK-NEXT: ParmVarDecl {{.*}} used s 'S'
+// CHECK-NEXT: CompoundStmt
+// CHECK-NEXT: CallExpr {{.*}} 'void'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(P)' <FunctionToPointerDecay>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'void (P)' lvalue Function {{.*}} 'useP' 'void (P)'
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'P' <DerivedToBase (P)>
+// CHECK-NEXT: ImplicitCastExpr {{.*}} 'S' <LValueToRValue>
+// CHECK-NEXT: DeclRefExpr {{.*}} 'S' lvalue ParmVar {{.*}} 's' 'S'
+void case6(S s) {
+ // derived to base conversion in argument passing
+ useP(s);
+}
+
+// CHECK-NOT: CXXConstructExpr
+// CHECK-NOT: CXXOperatorCallExpr
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl
new file mode 100644
index 0000000000000..5c6940c4d2e4d
--- /dev/null
+++ b/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl
@@ -0,0 +1,113 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
+
+struct P {
+ float a;
+};
+
+struct S : P {
+ double b;
+ int c[2];
+};
+
+// CHECK-LABEL: define hidden void @_Z5case11S(
+// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[SLOCAL:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[SLOCAL]], ptr align 1 [[S]], i32 20, i1 false)
+// CHECK-NEXT: ret void
+//
+void case1(S s) {
+ // struct initialization
+ S sLocal = s;
+}
+
+// CHECK-LABEL: define hidden void @_Z5case21S(
+// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[SLOCAL:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_S]], align 1
+// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[SLOCAL]], ptr align 1 [[S]], i32 20, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TMP]], ptr align 1 [[SLOCAL]], i32 20, i1 false)
+// CHECK-NEXT: ret void
+//
+void case2(S s) {
+ S sLocal;
+ // struct assignment
+ sLocal = s;
+}
+
+void useS(S s) {}
+
+// CHECK-LABEL: define hidden void @_Z5case31S(
+// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[BYVAL_TEMP:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[BYVAL_TEMP]], ptr align 1 [[S]], i32 20, i1 false)
+// CHECK-NEXT: call void @_Z4useS1S(ptr noundef dead_on_return [[BYVAL_TEMP]]) #[[ATTR2:[0-9]+]]
+// CHECK-NEXT: ret void
+//
+void case3(S s) {
+ // struct argument passing
+ useS(s);
+}
+
+// CHECK-LABEL: define hidden void @_Z5case41S(
+// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[PLOCAL:%.*]] = alloca [[STRUCT_P:%.*]], align 1
+// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TMP]], ptr align 1 [[S]], i32 20, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[PLOCAL]], ptr align 1 [[TMP]], i32 4, i1 false)
+// CHECK-NEXT: ret void
+//
+void case4(S s) {
+ // derived to base conversion in initialization
+ P pLocal = s;
+}
+
+// CHECK-LABEL: define hidden void @_Z5case51S(
+// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[P:%.*]] = alloca [[STRUCT_P:%.*]], align 1
+// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_P]], align 1
+// CHECK-NEXT: [[TMP1:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TMP1]], ptr align 1 [[S]], i32 20, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[P]], ptr align 1 [[TMP1]], i32 4, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TMP]], ptr align 1 [[P]], i32 4, i1 false)
+// CHECK-NEXT: ret void
+//
+void case5(S s) {
+ P p;
+ // derived to base conversion in assignment
+ p = s;
+}
+
+void useP(P p) {}
+
+// CHECK-LABEL: define hidden void @_Z5case61S(
+// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[AGG_TMP:%.*]] = alloca [[STRUCT_P:%.*]], align 1
+// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TMP]], ptr align 1 [[S]], i32 20, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AGG_TMP]], ptr align 1 [[TMP]], i32 4, i1 false)
+// CHECK-NEXT: call void @_Z4useP1P(ptr noundef dead_on_return [[AGG_TMP]]) #[[ATTR2]]
+// CHECK-NEXT: ret void
+//
+void case6(S s) {
+ // derived to base conversion in argument passing
+ useP(s);
+}
diff --git a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
index 7858fbcf68543..d1eb2f740dbf2 100644
--- a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
+++ b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
@@ -125,7 +125,7 @@ void case1() {
// CHECK-NEXT: ret void
void case2() {
- S AggTemp = (S)cbs;
+ S LocalS = (S)cbs;
}
// CHECK-LABEL: case3
@@ -135,8 +135,8 @@ void case3() {
// CHECK-NEXT: [[LocalT:%.*]] = alloca %struct.T, align 1
// CHECK-NEXT: [[LocalTCopy:%.*]] = alloca %struct.T, align 1
-// CHECK-NEXT: [[AggTemp:%.*]] = alloca %struct.S, align 1
-// CHECK-NEXT: [[AggTempCopy:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[LocalS:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[LocalSCopy:%.*]] = alloca %struct.S, align 1
// Check that constant to default address space copies the struct field by field
//
@@ -170,21 +170,70 @@ void case3() {
// Check that constant to default address space copies the struct field by field
//
-// CHECK-NEXT: [[Ptr_a1:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 0
+// CHECK-NEXT: [[Ptr_a1:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 0
// CHECK-NEXT: [[CbufLoad_a1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, align 4
// CHECK-NEXT: store <3 x float> [[CbufLoad_a1]], ptr [[Ptr_a1]], align 4
-// CHECK-NEXT: [[Ptr_b1:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 1
+// CHECK-NEXT: [[Ptr_b1:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 1
// CHECK-NEXT: [[CbufLoad_b1:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 16), align 8
// CHECK-NEXT: store double [[CbufLoad_b1]], ptr [[Ptr_b1]], align 8
-// CHECK-NEXT: [[Ptr_c1:%.*]] = getelementptr inbounds %struct.S, ptr [[AggTemp]], i32 0, i32 2
+// CHECK-NEXT: [[Ptr_c1:%.*]] = getelementptr inbounds %struct.S, ptr [[LocalS]], i32 0, i32 2
// CHECK-NEXT: [[CbufLoad_c1:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbt, {{i32|i64}} 32), align 4
// CHECK-NEXT: store <4 x float> [[CbufLoad_c1]], ptr [[Ptr_c1]], align 4
- S AggTemp = cbt.s;
+ S localS = cbt.s;
// Check that default to default address space copy uses memcpy
//
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[AggTempCopy]], ptr align 1 [[AggTemp]], {{i32|i64}} 36, i1 false)
- S AggTempCopy = AggTemp;
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[LocalSCopy]], ptr align 1 [[LocalS]], {{i32|i64}} 36, i1 false)
+ S localSCopy = localS;
+
+// CHECK-NEXT: ret void
+}
+
+// CHECK-LABEL: case4
+// CHECK-NEXT: entry:
+// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+void case4() {
+
+// CHECK-NEXT: [[LocalP1:%.*]] = alloca %struct.P, align 1
+// CHECK-NEXT: [[Tmp0:%.*]] = alloca %struct.S, align 1
+// CHECK-NEXT: [[LocalP2:%.*]] = alloca %struct.P, align 1
+// CHECK-NEXT: [[Tmp1:%.*]] = alloca %struct.P, align 1
+// CHECK-NEXT: [[Tmp2:%.*]] = alloca %struct.S, align 1
+
+// CHECK-NEXT: [[Tmp0Ptr_a1:%.*]] = getelementptr inbounds %struct.S, ptr [[Tmp0]], i32 0, i32 0
+// CHECK-NEXT: [[CbufLoad_a1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
+// CHECK-NEXT: store <3 x float> [[CbufLoad_a1]], ptr [[Tmp0Ptr_a1]], align 4
+
+// CHECK-NEXT: [[Tmp0Ptr_b1:%.*]] = getelementptr inbounds %struct.S, ptr [[Tmp0]], i32 0, i32 1
+// CHECK-NEXT: [[CbufLoad_b1:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 8
+// CHECK-NEXT: store double [[CbufLoad_b1]], ptr [[Tmp0Ptr_b1]], align 8
+
+// CHECK-NEXT: [[Tmp0Ptr_c1:%.*]] = getelementptr inbounds %struct.S, ptr [[Tmp0]], i32 0, i32 2
+// CHECK-NEXT: [[CbufLoad_c1:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 32), align 4
+// CHECK-NEXT: store <4 x float> [[CbufLoad_c1]], ptr [[Tmp0Ptr_c1]], align 4
+
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[LocalP1]], ptr align 1 [[Tmp0]], {{i32|i64}} 12, i1 false)
+
+ // Derived to base conversion in initialization. Size of S in memory layout is 36 bytes and
+ // size of P is 12 bytes. The memcpy should only copy the 12 bytes of P.
+ P LocalP1 = cbs;
+
+// CHECK-NEXT: [[Tmp2Ptr_a1:%.*]] = getelementptr inbounds %struct.S, ptr [[Tmp2]], i32 0, i32 0
+// CHECK-NEXT: [[CbufLoad_a1:%.*]] = load <3 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, align 4
+// CHECK-NEXT: store <3 x float> [[CbufLoad_a1]], ptr [[Tmp2Ptr_a1]], align 4
+// CHECK-NEXT: [[Tmp2Ptr_b1:%.*]] = getelementptr inbounds %struct.S, ptr [[Tmp2]], i32 0, i32 1
+// CHECK-NEXT: [[CbufLoad_b1:%.*]] = load double, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 16), align 8
+// CHECK-NEXT: store double [[CbufLoad_b1]], ptr [[Tmp2Ptr_b1]], align 8
+// CHECK-NEXT: [[Tmp2Ptr_c1:%.*]] = getelementptr inbounds %struct.S, ptr [[Tmp2]], i32 0, i32 2
+// CHECK-NEXT: [[CbufLoad_c1:%.*]] = load <4 x float>, ptr addrspace([[CONST_ADDR_SPACE]]) getelementptr inbounds nuw (i8, ptr addrspace([[CONST_ADDR_SPACE]]) @cbs, {{i32|i64}} 32), align 4
+// CHECK-NEXT: store <4 x float> [[CbufLoad_c1]], ptr [[Tmp2Ptr_c1]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[LocalP2]], ptr align 1 [[Tmp2]], {{i32|i64}} 12, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.{{i32|i64}}(ptr align 1 [[Tmp1]], ptr align 1 [[LocalP2]], {{i32|i64}} 12, i1 false)
+
+ // Derived to base conversion in assignment. Size of S in memory layout is 36 bytes and
+ // size of P is 12 bytes. The memcpy should only copy the 12 bytes of P.
+ P LocalP2;
+ LocalP2 = cbs;
// CHECK-NEXT: ret void
}
>From 13dc39199b3f00e01ade4b4a00523c50ce14b564 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 15 May 2026 13:27:37 -0700
Subject: [PATCH 09/22] fix typos in comments
---
clang/include/clang/AST/DeclCXX.h | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index 2a73e56f2465f..dd993187f9244 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -764,7 +764,7 @@ class CXXRecordDecl : public RecordDecl {
// have ctors, dtors, or overloaded operators, while implicit built-in
// HLSL records such as resource classes can. It would be nice to use the
// isImplicit() methods to determine that, but this flag is not propagated
- // to template-instanticated classes.
+ // to template-instantiated classes.
//
/// Determines whether this class has any user provided special members.
bool hasUserProvidedSpecialMembers() const {
@@ -778,7 +778,7 @@ class CXXRecordDecl : public RecordDecl {
///
/// This value is used for lazy creation of default constructors.
bool needsImplicitDefaultConstructor() const {
- // In HLSL, only built-in records like resources classes can have
+ // In HLSL, only built-in records like resource classes can have
// constructors and overloadable operators.
if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
return false;
@@ -816,7 +816,7 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// constructor to be lazily declared.
bool needsImplicitCopyConstructor() const {
- // In HLSL, only built-in records like resources classes can have
+ // In HLSL, only built-in records like resource classes can have
// constructors and overloadable operators.
if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
return false;
@@ -914,7 +914,7 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class should get an implicit move
/// constructor or if any existing special member function inhibits this.
bool needsImplicitMoveConstructor() const {
- // In HLSL, only built-in records like resources classes can have
+ // In HLSL, only built-in records like resource classes can have
// constructors and overloadable operators.
if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
return false;
@@ -952,7 +952,7 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// assignment operator to be lazily declared.
bool needsImplicitCopyAssignment() const {
- // In HLSL, only built-in records like resources classes can have
+ // In HLSL, only built-in records like resource classes can have
// constructors and overloadable operators.
if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
return false;
@@ -1015,7 +1015,7 @@ class CXXRecordDecl : public RecordDecl {
/// assignment operator or if any existing special member function inhibits
/// this.
bool needsImplicitMoveAssignment() const {
- // In HLSL, only built-in records like resources classes can have
+ // In HLSL, only built-in records like resource classes can have
// constructors and overloadable operators.
if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
return false;
>From 12cf8f4de5ae17aa8420aa4138037e9e8542b41a Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 15 May 2026 13:40:33 -0700
Subject: [PATCH 10/22] Update AST test to change that default and move
constructors and move assignment do not exist.
---
clang/test/AST/HLSL/StructPassing-AST.hlsl | 22 ++++++++--------------
1 file changed, 8 insertions(+), 14 deletions(-)
diff --git a/clang/test/AST/HLSL/StructPassing-AST.hlsl b/clang/test/AST/HLSL/StructPassing-AST.hlsl
index 8b6b4bbe9a824..8ae83fa89b390 100644
--- a/clang/test/AST/HLSL/StructPassing-AST.hlsl
+++ b/clang/test/AST/HLSL/StructPassing-AST.hlsl
@@ -2,13 +2,10 @@
// CHECK: CXXRecordDecl {{.*}} struct P definition
// CHECK-NEXT: DefinitionData aggregate standard_layout trivially_copyable pod literal can_const_default_init
-// CHECK-NEXT: DefaultConstructor
-// CHECK-NEXT: CopyConstructor simple trivial implicit_has_const_param
-// CHECK-NEXT: MoveConstructor
-// CHECK-NEXT: CopyAssignment simple trivial implicit_has_const_param
-// CHECK-NEXT: MoveAssignment
-// CHECK-NEXT: Destructor simple irrelevant trivial needs_implicit
-// CHECK-NEXT: PackedAttr
+// CHECK-NOT: DefaultConstructor {{.*}} exists
+// CHECK-NOT: MoveConstructor {{.*}} exists
+// CHECK-NOT: MoveAssignment {{.*}} exists
+// CHECK: PackedAttr
// CHECK-NEXT: CXXRecordDecl {{.*}} struct P
// CHECK-NEXT: FieldDecl {{.*}} a 'float'
// CHECK-NOT: CXXConstructorDecl
@@ -19,13 +16,10 @@ struct P {
// CHECK: CXXRecordDecl {{.*}} struct S definition
// CHECK-NEXT: DefinitionData aggregate trivially_copyable literal can_const_default_init
-// CHECK-NEXT: DefaultConstructor
-// CHECK-NEXT: CopyConstructor simple trivial
-// CHECK-NEXT: MoveConstructor needs_overload_resolution
-// CHECK-NEXT: CopyAssignment simple trivial
-// CHECK-NEXT: MoveAssignment needs_overload_resolution
-// CHECK-NEXT: Destructor simple irrelevant trivial needs_implicit
-// CHECK-NEXT: public 'P'
+// CHECK-NOT: DefaultConstructor {{.*}} exists
+// CHECK-NOT: MoveConstructor {{.*}} exists
+// CHECK-NOT: MoveAssignment {{.*}} exists
+// CHECK: public 'P'
// CHECK-NEXT: PackedAttr
// CHECK-NEXT: CXXRecordDecl {{.*}} implicit struct S
// CHECK-NEXT: FieldDecl {{.*}} b 'double'
>From 60ad20aa0ea27b2af3d2ecd109c47fff8bbf2a21 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 15 May 2026 14:09:05 -0700
Subject: [PATCH 11/22] clang-format
---
clang/include/clang/AST/DeclCXX.h | 2 +-
clang/lib/CodeGen/CGHLSLRuntime.cpp | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index 9cdc0613dbdab..36cd8f519ea34 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -1026,7 +1026,7 @@ class CXXRecordDecl : public RecordDecl {
!hasUserDeclaredMoveConstructor() &&
!hasUserDeclaredDestructor() &&
(!isLambda() || lambdaIsDefaultConstructibleAndAssignable());
- }
+ }
/// Determine whether we need to eagerly declare a move assignment
/// operator for this class.
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.cpp b/clang/lib/CodeGen/CGHLSLRuntime.cpp
index 538dda8ed0881..555835a7966fc 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.cpp
+++ b/clang/lib/CodeGen/CGHLSLRuntime.cpp
@@ -1086,7 +1086,7 @@ void CGHLSLRuntime::emitEntryFunction(const FunctionDecl *FD,
ParamType = CGM.getTypes().ConvertType(PD->getType());
else
ParamType = Param.getType();
-
+
auto AttrBegin = PD->specific_attr_begin<HLSLAppliedSemanticAttr>();
auto AttrEnd = PD->specific_attr_end<HLSLAppliedSemanticAttr>();
auto Result =
>From 0a34df32a483bd43c234e9afe6798156fbcda9ca Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 15 May 2026 18:35:59 -0700
Subject: [PATCH 12/22] Update tests after merge
---
clang/test/AST/HLSL/ConstantBuffers-AST-error.hlsl | 2 +-
clang/test/CodeGenHLSL/cbuffer_copy_layout.hlsl | 1 -
.../CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl | 3 ++-
clang/test/SemaHLSL/Resources/ConstantBuffers.hlsl | 2 +-
4 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/clang/test/AST/HLSL/ConstantBuffers-AST-error.hlsl b/clang/test/AST/HLSL/ConstantBuffers-AST-error.hlsl
index 3e2d8075a6569..4f9d60c741f90 100644
--- a/clang/test/AST/HLSL/ConstantBuffers-AST-error.hlsl
+++ b/clang/test/AST/HLSL/ConstantBuffers-AST-error.hlsl
@@ -19,6 +19,6 @@ void main() {
// CHECK: error: no viable constructor copying variable of type 'const hlsl_constant S'
S s2 = cb;
- // CHECK: error: no viable conversion from 'ConstantBuffer<S>' to 'const S'
+ // CHECK: error: assigning to 'S' from incompatible type 'ConstantBuffer<S>'
s = cb;
}
diff --git a/clang/test/CodeGenHLSL/cbuffer_copy_layout.hlsl b/clang/test/CodeGenHLSL/cbuffer_copy_layout.hlsl
index ed1fe3ac0014e..022844284f4ba 100644
--- a/clang/test/CodeGenHLSL/cbuffer_copy_layout.hlsl
+++ b/clang/test/CodeGenHLSL/cbuffer_copy_layout.hlsl
@@ -16,7 +16,6 @@ ConstantBuffer<S> cb;
[numthreads(1,1,1)]
void main() {
- // CHECK: error: no matching constructor for initialization of 'S'
S l1 = s_cb;
// CHECK: error: no viable constructor copying variable of type 'const hlsl_constant S'
diff --git a/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl b/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
index 4e1c1b7b55984..9065d469e5522 100644
--- a/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
+++ b/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
@@ -19,6 +19,7 @@ void main(unsigned GI : SV_GroupIndex) {
// CHECK: define void @main()
// CHECK: %[[TMP:.*]] = alloca %struct.S, align 1
+ // SPV: %[[TMPCAST:.*]] = addrspacecast ptr %[[TMP]] to ptr addrspace(11)
// DXIL: %[[INPTR:.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_i32_0_0t.i32(target("dx.RawBuffer", i32, 0, 0) %{{.*}}, i32 %{{.*}})
// SPV: %[[INPTR:.*]] = call noundef align 4 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0i32_12_0t.i32(target("spirv.VulkanBuffer", [0 x i32], 12, 0) %{{.*}}, i32 %{{.*}})
@@ -39,7 +40,7 @@ void main(unsigned GI : SV_GroupIndex) {
// SPV: %[[OUTPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
// SPV: %[[INPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
// SPV: call void @llvm.memcpy.p11.p11.i64(ptr addrspace(11) align 1 %[[OUTPTR]], ptr addrspace(11) align 1 %[[INPTR]], i64 4, i1 false)
- // SPV: call void @llvm.memcpy.p0.p11.i64(ptr align 1 %[[TMP]], ptr addrspace(11) align 1 %[[OUTPTR]], i64 4, i1 false)
+ // SPV: call void @llvm.memcpy.p11.p11.i64(ptr addrspace(11) align 1 %[[TMPCAST]], ptr addrspace(11) align 1 %[[OUTPTR]], i64 4, i1 false)
// DXIL: %[[OUTPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
// DXIL: %[[INPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
diff --git a/clang/test/SemaHLSL/Resources/ConstantBuffers.hlsl b/clang/test/SemaHLSL/Resources/ConstantBuffers.hlsl
index 0ef3ada50c988..6dc566a58b5e4 100644
--- a/clang/test/SemaHLSL/Resources/ConstantBuffers.hlsl
+++ b/clang/test/SemaHLSL/Resources/ConstantBuffers.hlsl
@@ -1,6 +1,6 @@
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -x hlsl -finclude-default-header -fsyntax-only -verify %s
-struct S { // expected-note 3 {{candidate constructor}}
+struct S {
float a;
int b;
};
>From de727fbdd9dbed318d784555db999b6fcb26191d Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Mon, 18 May 2026 16:03:48 -0700
Subject: [PATCH 13/22] Update test after merge
---
.../CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl b/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
index 9065d469e5522..4e1c1b7b55984 100644
--- a/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
+++ b/clang/test/CodeGenHLSL/resources/StructuredBuffers-subscripts.hlsl
@@ -19,7 +19,6 @@ void main(unsigned GI : SV_GroupIndex) {
// CHECK: define void @main()
// CHECK: %[[TMP:.*]] = alloca %struct.S, align 1
- // SPV: %[[TMPCAST:.*]] = addrspacecast ptr %[[TMP]] to ptr addrspace(11)
// DXIL: %[[INPTR:.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_i32_0_0t.i32(target("dx.RawBuffer", i32, 0, 0) %{{.*}}, i32 %{{.*}})
// SPV: %[[INPTR:.*]] = call noundef align 4 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0i32_12_0t.i32(target("spirv.VulkanBuffer", [0 x i32], 12, 0) %{{.*}}, i32 %{{.*}})
@@ -40,7 +39,7 @@ void main(unsigned GI : SV_GroupIndex) {
// SPV: %[[OUTPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
// SPV: %[[INPTR:.*]] = call noundef align 1 dereferenceable(4) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0s_struct.Ss_12_1t.i32(target("spirv.VulkanBuffer", [0 x %struct.S], 12, 1) %{{.*}}, i32 %{{.*}})
// SPV: call void @llvm.memcpy.p11.p11.i64(ptr addrspace(11) align 1 %[[OUTPTR]], ptr addrspace(11) align 1 %[[INPTR]], i64 4, i1 false)
- // SPV: call void @llvm.memcpy.p11.p11.i64(ptr addrspace(11) align 1 %[[TMPCAST]], ptr addrspace(11) align 1 %[[OUTPTR]], i64 4, i1 false)
+ // SPV: call void @llvm.memcpy.p0.p11.i64(ptr align 1 %[[TMP]], ptr addrspace(11) align 1 %[[OUTPTR]], i64 4, i1 false)
// DXIL: %[[OUTPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
// DXIL: %[[INPTR:.*]] = call noundef nonnull align 1 dereferenceable(4) ptr @llvm.dx.resource.getpointer.p0.tdx.RawBuffer_s_struct.Ss_1_0t.i32(target("dx.RawBuffer", %struct.S, 1, 0) %{{.*}}, i32 %{{.*}})
>From bd00ba0c96cd1ebe3913cc8b6d3aeb08cee394f0 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 21 May 2026 12:13:21 -0700
Subject: [PATCH 14/22] code review feedback - HLSL-specific isPODType, update
overloading binop check
---
clang/include/clang/AST/TypeBase.h | 5 +
clang/include/clang/Sema/SemaHLSL.h | 2 +
clang/lib/AST/Type.cpp | 20 +++
clang/lib/CodeGen/CGDecl.cpp | 8 +-
clang/lib/Sema/SemaExpr.cpp | 21 +--
clang/lib/Sema/SemaHLSL.cpp | 11 ++
clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak | 160 ++++++++++++++++++
.../this-assignment-overload.hlsl.bak | 58 +++++++
8 files changed, 264 insertions(+), 21 deletions(-)
create mode 100644 clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak
create mode 100644 clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak
diff --git a/clang/include/clang/AST/TypeBase.h b/clang/include/clang/AST/TypeBase.h
index c64eee11fd91e..779a4eba3468d 100644
--- a/clang/include/clang/AST/TypeBase.h
+++ b/clang/include/clang/AST/TypeBase.h
@@ -1111,6 +1111,11 @@ class QualType {
/// CXXRecordDecl::isCXX11StandardLayout, this takes DRs into account.
bool isCXX11PODType(const ASTContext &Context) const;
+ /// Return true if this is a POD type according to the HLSL rules.
+ /// User-defined records in HLSL do not have default constructors, which
+ /// is a POD requirement for C++.
+ bool isHLSLPODType(const ASTContext &Context) const;
+
/// Return true if this is a trivial type per (C++0x [basic.types]p9)
bool isTrivialType(const ASTContext &Context) const;
diff --git a/clang/include/clang/Sema/SemaHLSL.h b/clang/include/clang/Sema/SemaHLSL.h
index e65de5d4aa4c3..02304ba0891f8 100644
--- a/clang/include/clang/Sema/SemaHLSL.h
+++ b/clang/include/clang/Sema/SemaHLSL.h
@@ -138,6 +138,8 @@ class SemaHLSL : public SemaBase {
bool CheckResourceBinOp(BinaryOperatorKind Opc, Expr *LHSExpr, Expr *RHSExpr,
SourceLocation Loc);
+ bool canHaveOverloadedBinOp(QualType Ty, BinaryOperatorKind Opc);
+
QualType handleVectorBinOpConversion(ExprResult &LHS, ExprResult &RHS,
QualType LHSType, QualType RHSType,
bool IsCompAssign);
diff --git a/clang/lib/AST/Type.cpp b/clang/lib/AST/Type.cpp
index 96a398aa21dad..d4504e6b04f8c 100644
--- a/clang/lib/AST/Type.cpp
+++ b/clang/lib/AST/Type.cpp
@@ -2786,6 +2786,10 @@ QualType Type::getRVVEltType(const ASTContext &Ctx) const {
}
bool QualType::isPODType(const ASTContext &Context) const {
+ // HLSL has a more relaxed definition of POD than C++11.
+ if (Context.getLangOpts().HLSL)
+ return isHLSLPODType(Context);
+
// C++11 has a more relaxed definition of POD.
if (Context.getLangOpts().CPlusPlus11)
return isCXX11PODType(Context);
@@ -3290,6 +3294,22 @@ bool QualType::isCXX11PODType(const ASTContext &Context) const {
return false;
}
+bool QualType::isHLSLPODType(const ASTContext &Context) const {
+ if (isCXX11PODType(Context))
+ return true;
+
+ const Type *BaseTy = getTypePtr()->getBaseElementTypeUnsafe();
+ if (const auto *RD =
+ dyn_cast_or_null<CXXRecordDecl>(BaseTy->getAsRecordDecl())) {
+ // User-defined records in HLSL do not have constructors or copy/assignment
+ // operators. They are still considered POD.
+ if (!RD->hasUserProvidedSpecialMembers() || RD->isTrivial() ||
+ RD->isStandardLayout())
+ return true;
+ }
+ return false;
+}
+
bool Type::isNothrowT() const {
if (const auto *RD = getAsCXXRecordDecl()) {
IdentifierInfo *II = RD->getIdentifier();
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index feeedc94de179..7608f8cb6fc7a 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -1525,14 +1525,10 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
// isConstantInitializer produces wrong answers for structs with
// reference or bitfield members, and a few other cases, and checking
// for POD-ness protects us from some of these.
- QualType BaseTy = getContext().getBaseElementType(Ty);
if (D.getInit() && (Ty->isArrayType() || Ty->isRecordType()) &&
(D.isConstexpr() ||
- ((Ty.isPODType(getContext()) || BaseTy->isObjCObjectPointerType() ||
- // If HLSL, check if it's a constant initializer anyway because
- // POD-ness will no longer be true for user defined structs
- // (since they do not have constructors).
- (getLangOpts().HLSL && BaseTy->isRecordType())) &&
+ ((Ty.isPODType(getContext()) ||
+ getContext().getBaseElementType(Ty)->isObjCObjectPointerType()) &&
D.getInit()->isConstantInitializer(getContext())))) {
// If the variable's a const type, and it's neither an NRVO
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 50fcff9001f88..7d28fec1dbcc2 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -16125,21 +16125,12 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
}
if (getLangOpts().CPlusPlus) {
- // Otherwise, build an overloaded op if either expression is type-dependent
- // or has an overloadable type.
- // In HLSL, user-defined structs/classes do not have ctors or
- // overloadable operators.
- QualType LHSTy = LHSExpr->getType();
- QualType RHSTy = RHSExpr->getType();
- bool IsLHSNonOverloadableHLSLType =
- getLangOpts().HLSL && LHSTy->isRecordType() &&
- !LHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers();
- bool IsRHSNonOverloadableHLSLType =
- getLangOpts().HLSL && RHSTy->isRecordType() &&
- !RHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers();
- if (LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent() ||
- (LHSTy->isOverloadableType() && !IsLHSNonOverloadableHLSLType) ||
- (RHSTy->isOverloadableType() && !IsRHSNonOverloadableHLSLType))
+ if ((LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent() ||
+ LHSExpr->getType()->isOverloadableType() ||
+ RHSExpr->getType()->isOverloadableType()) &&
+ (!getLangOpts().HLSL ||
+ HLSL().canHaveOverloadedBinOp(LHSExpr->getType(), Opc) ||
+ HLSL().canHaveOverloadedBinOp(RHSExpr->getType(), Opc)))
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
}
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index ad87e22e49343..c201f24cf5049 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -5745,6 +5745,17 @@ bool SemaHLSL::CheckResourceBinOp(BinaryOperatorKind Opc, Expr *LHSExpr,
return true;
}
+// Returns true if the given type can have an overload of the given
+// binary operator.
+bool SemaHLSL::canHaveOverloadedBinOp(QualType LHSTy, BinaryOperatorKind Opc) {
+ CXXRecordDecl *RD = LHSTy->getAsCXXRecordDecl();
+ if (!RD)
+ return true;
+ // hasUserProvidedSpecialMembers() should be true only for HLSL built-in
+ // records like resources.
+ return RD->hasUserProvidedSpecialMembers() || Opc != BO_Assign;
+}
+
// Walks though the global variable declaration, collects all resource binding
// requirements and adds them to Bindings
void SemaHLSL::collectResourceBindingsOnVarDecl(VarDecl *VD) {
diff --git a/clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak b/clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak
new file mode 100644
index 0000000000000..a17e0ae6038b1
--- /dev/null
+++ b/clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak
@@ -0,0 +1,160 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
+
+// CHECK: [[ConstInit_fn3_s:@.*]] = private unnamed_addr constant { <4 x i1>, [12 x i8], float } { <4 x i1> <i1 true, i1 false, i1 true, i1 false>, [12 x i8] undef, float 1.000000e+00 }, align 1
+// CHECK: [[ConstInit_fn4_Arr:@.*]] = private unnamed_addr constant [2 x <4 x i1>] [<4 x i1> splat (i1 true), <4 x i1> zeroinitializer], align 4
+// CHECK: [[ConstInit_fn6_s:@.*]] = private unnamed_addr constant { <4 x i1>, [12 x i8], float } { <4 x i1> <i1 true, i1 false, i1 true, i1 false>, [12 x i8] undef, float 1.000000e+00 }, align 1
+// CHECK: [[ConstInit_fn7_Arr:@.*]] = private unnamed_addr constant [2 x <4 x i1>] [<4 x i1> splat (i1 true), <4 x i1> zeroinitializer], align 4
+
+
+struct S {
+ bool2x2 bM;
+ float f;
+};
+
+// CHECK-LABEL: define hidden noundef i1 @_Z3fn1v(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RETVAL:%.*]] = alloca i1, align 4
+// CHECK-NEXT: [[B:%.*]] = alloca [2 x <2 x i32>], align 4
+// CHECK-NEXT: store <4 x i32> splat (i32 1), ptr [[B]], align 4
+// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[B]], align 4
+// CHECK-NEXT: [[MATRIXEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 0
+// CHECK-NEXT: store i32 [[MATRIXEXT]], ptr [[RETVAL]], align 4
+// CHECK-NEXT: [[TMP1:%.*]] = load i1, ptr [[RETVAL]], align 4
+// CHECK-NEXT: ret i1 [[TMP1]]
+//
+bool fn1() {
+ bool2x2 B = {true,true,true,true};
+ return B[0][0];
+}
+
+// CHECK-LABEL: define hidden noundef <4 x i1> @_Z3fn2b(
+// CHECK-SAME: i1 noundef [[V:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x i1>, align 4
+// CHECK-NEXT: [[V_ADDR:%.*]] = alloca i32, align 4
+// CHECK-NEXT: [[A:%.*]] = alloca [2 x <2 x i32>], align 4
+// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[V]] to i32
+// CHECK-NEXT: store i32 [[STOREDV]], ptr [[V_ADDR]], align 4
+// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[V_ADDR]], align 4
+// CHECK-NEXT: [[LOADEDV:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK-NEXT: [[VECINIT:%.*]] = insertelement <4 x i1> poison, i1 [[LOADEDV]], i32 0
+// CHECK-NEXT: [[VECINIT1:%.*]] = insertelement <4 x i1> [[VECINIT]], i1 true, i32 2
+// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[V_ADDR]], align 4
+// CHECK-NEXT: [[LOADEDV2:%.*]] = icmp ne i32 [[TMP1]], 0
+// CHECK-NEXT: [[VECINIT3:%.*]] = insertelement <4 x i1> [[VECINIT1]], i1 [[LOADEDV2]], i32 1
+// CHECK-NEXT: [[VECINIT4:%.*]] = insertelement <4 x i1> [[VECINIT3]], i1 false, i32 3
+// CHECK-NEXT: [[TMP2:%.*]] = zext <4 x i1> [[VECINIT4]] to <4 x i32>
+// CHECK-NEXT: store <4 x i32> [[TMP2]], ptr [[A]], align 4
+// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[A]], align 4
+// CHECK-NEXT: store <4 x i32> [[TMP3]], ptr [[RETVAL]], align 4
+// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i1>, ptr [[RETVAL]], align 4
+// CHECK-NEXT: ret <4 x i1> [[TMP4]]
+//
+bool2x2 fn2(bool V) {
+ bool2x2 A = {V, true, V, false};
+ return A;
+}
+
+// CHECK-LABEL: define hidden noundef i1 @_Z3fn3v(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RETVAL:%.*]] = alloca i1, align 4
+// CHECK-NEXT: [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[S]], ptr align 1 [[ConstInit_fn3_s]], i32 20, i1 false)
+// CHECK-NEXT: [[BM:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[S]], i32 0, i32 0
+// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[BM]], align 1
+// CHECK-NEXT: [[MATRIXEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 0
+// CHECK-NEXT: store i32 [[MATRIXEXT]], ptr [[RETVAL]], align 4
+// CHECK-NEXT: [[TMP1:%.*]] = load i1, ptr [[RETVAL]], align 4
+// CHECK-NEXT: ret i1 [[TMP1]]
+//
+bool fn3() {
+ S s = {{true,true,false,false}, 1.0};
+ return s.bM[0][0];
+}
+
+// CHECK-LABEL: define hidden noundef i1 @_Z3fn4v(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RETVAL:%.*]] = alloca i1, align 4
+// CHECK-NEXT: [[ARR:%.*]] = alloca [2 x [2 x <2 x i32>]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 %Arr, ptr align 4 [[ConstInit_fn4_Arr]], i32 8, i1 false)
+// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x [2 x <2 x i32>]], ptr [[ARR]], i32 0, i32 0
+// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[ARRAYIDX]], align 4
+// CHECK-NEXT: [[MATRIXEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 1
+// CHECK-NEXT: store i32 [[MATRIXEXT]], ptr [[RETVAL]], align 4
+// CHECK-NEXT: [[TMP1:%.*]] = load i1, ptr [[RETVAL]], align 4
+// CHECK-NEXT: ret i1 [[TMP1]]
+//
+bool fn4() {
+ bool2x2 Arr[2] = {{true,true,true,true}, {false,false,false,false}};
+ return Arr[0][1][0];
+}
+
+// CHECK-LABEL: define hidden void @_Z3fn5v(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[M:%.*]] = alloca [2 x <2 x i32>], align 4
+// CHECK-NEXT: store <4 x i32> splat (i32 1), ptr [[M]], align 4
+// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <4 x i32>, ptr [[M]], i32 0, i32 3
+// CHECK-NEXT: store i32 0, ptr [[TMP0]], align 4
+// CHECK-NEXT: ret void
+//
+void fn5() {
+ bool2x2 M = {true,true,true,true};
+ M[1][1] = false;
+}
+
+// CHECK-LABEL: define hidden void @_Z3fn6v(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[V:%.*]] = alloca i32, align 4
+// CHECK-NEXT: [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 1
+// CHECK-NEXT: store i32 0, ptr [[V]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[S]], ptr align 1 [[ConstInit_fn6_s]], i32 20, i1 false)
+// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[V]], align 4
+// CHECK-NEXT: [[LOADEDV:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK-NEXT: [[BM:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[S]], i32 0, i32 0
+// CHECK-NEXT: [[TMP1:%.*]] = zext i1 [[LOADEDV]] to i32
+// CHECK-NEXT: [[TMP2:%.*]] = getelementptr <4 x i32>, ptr [[BM]], i32 0, i32 1
+// CHECK-NEXT: store i32 [[TMP1]], ptr [[TMP2]], align 4
+// CHECK-NEXT: ret void
+//
+void fn6() {
+ bool V = false;
+ S s = {{true,true,false,false}, 1.0};
+ s.bM[1][0] = V;
+}
+
+// CHECK-LABEL: define hidden void @_Z3fn7v(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[ARR:%.*]] = alloca [2 x [2 x <2 x i32>]], align 4
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARR]], ptr align 4 [[ConstInit_fn7_Arr]], i32 8, i1 false)
+// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x [2 x <2 x i32>]], ptr [[ARR]], i32 0, i32 0
+// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <4 x i32>, ptr [[ARRAYIDX]], i32 0, i32 1
+// CHECK-NEXT: store i32 0, ptr [[TMP0]], align 4
+// CHECK-NEXT: ret void
+//
+void fn7() {
+ bool2x2 Arr[2] = {{true,true,true,true}, {false,false,false,false}};
+ Arr[0][1][0] = false;
+}
+
+// CHECK-LABEL: define hidden noundef <16 x i1> @_Z3fn8u11matrix_typeILm4ELm4EbE(
+// CHECK-SAME: <16 x i1> noundef [[M:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: [[RETVAL:%.*]] = alloca <16 x i1>, align 4
+// CHECK-NEXT: [[M_ADDR:%.*]] = alloca [4 x <4 x i32>], align 4
+// CHECK-NEXT: [[TMP0:%.*]] = zext <16 x i1> [[M]] to <16 x i32>
+// CHECK-NEXT: store <16 x i32> [[TMP0]], ptr [[M_ADDR]], align 4
+// CHECK-NEXT: [[TMP1:%.*]] = load <16 x i32>, ptr [[M_ADDR]], align 4
+// CHECK-NEXT: store <16 x i32> [[TMP1]], ptr [[RETVAL]], align 4
+// CHECK-NEXT: [[TMP2:%.*]] = load <16 x i1>, ptr [[RETVAL]], align 4
+// CHECK-NEXT: ret <16 x i1> [[TMP2]]
+//
+bool4x4 fn8(bool4x4 m) {
+ return m;
+}
diff --git a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak
new file mode 100644
index 0000000000000..7b08f7b96c624
--- /dev/null
+++ b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak
@@ -0,0 +1,58 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -disable-llvm-passes -o - -std=hlsl202x %s | FileCheck %s
+
+struct Pair {
+ int First;
+ int Second;
+ int getFirst() {
+ Pair Another = {5, 10};
+ this = Another;
+ return this.First;
+ }
+ int getSecond() {
+ this = {0, 123};
+ return Second;
+ }
+ void operator=(Pair P) {
+ First = P.First;
+ Second = 2;
+ }
+};
+[numthreads(1, 1, 1)]
+void main() {
+ Pair Vals = {1, 2};
+ Vals.First = Vals.getFirst();
+ Vals.Second = Vals.getSecond();
+}
+
+// This test makes a probably safe assumption that HLSL 202x includes operator overloading for assignment operators.
+// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair8getFirstEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: %Another = alloca %struct.Pair, align 1
+// CHECK-NEXT: [[AggTmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr %Another, i32 0, i32 0
+// CHECK-NEXT: store i32 5, ptr [[First]], align 1
+// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr %Another, i32 0, i32 1
+// CHECK-NEXT: store i32 10, ptr [[Second]], align 1
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AggTmp]], ptr align 1 %Another, i32 8, i1 false)
+// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[AggTmp]])
+// CHECK-NEXT: [[First2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
+// CHECK-NEXT: %0 = load i32, ptr [[First2]], align 1
+// CHECK-NEXT: ret i32 %0
+
+// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair9getSecondEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
+// CHECK-NEXT:entry:
+// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
+// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 0
+// CHECK-NEXT: store i32 0, ptr [[First]], align 1
+// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 1
+// CHECK-NEXT: store i32 123, ptr [[Second]], align 1
+// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
+// CHECK-NEXT: [[Second2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
+// CHECK-NEXT: %0 = load i32, ptr [[Second2]], align 1
+// CHECK-NEXT: ret i32 %0
>From 91cf4b26bad70f23561335e2c43ff5d521422bb1 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Tue, 26 May 2026 16:31:43 -0700
Subject: [PATCH 15/22] Update tests after merge
---
.../BasicFeatures/ArrayElementwiseCast.hlsl | 1 +
clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl | 3 ++-
.../BasicFeatures/StructElementwiseCast.hlsl | 2 ++
.../test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl | 6 ++++++
.../BasicFeatures/VectorElementwiseCast.hlsl | 2 ++
.../CodeGenHLSL/resources/cbuffer_struct_passing.hlsl | 10 +++++-----
clang/test/CodeGenHLSL/this-assignment-overload.hlsl | 8 ++++----
7 files changed, 22 insertions(+), 10 deletions(-)
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
index 287973a7f7764..b8611914efb29 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
@@ -157,6 +157,7 @@ struct Derived : BFields {
// flatten from a derived struct with bitfields
// CHECK-LABEL: call8
// CHECK-NEXT: entry:
+// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[DIndirectAddr:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[A:%.*]] = alloca [4 x i32], align 4
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl
index 5873f70319de9..183bfb6f3d87a 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl
@@ -1181,7 +1181,7 @@ void case24() {
// CHECK-SAME: ptr noundef dead_on_return [[ED:%.*]], ptr noundef dead_on_return [[UD:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
- // CHECK-NEXT: [[ED_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
+// CHECK-NEXT: [[ED_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[UD_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[TI1:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1
// CHECK-NEXT: [[TI2:%.*]] = alloca [[STRUCT_TWOINTS]], align 1
@@ -1199,6 +1199,7 @@ void case25(EmptyDerived ED, UnnamedDerived UD) {
// CHECK-LABEL: define hidden void @_Z6case267TwoInts(
// CHECK-SAME: ptr noundef dead_on_return [[TI:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[TI_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[F:%.*]] = alloca <4 x float>, align 4
// CHECK-NEXT: [[F2:%.*]] = alloca <3 x float>, align 4
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl
index 3b7111ee2fd02..e02491d71cc72 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl
@@ -153,6 +153,7 @@ struct Derived : BFields {
// Derived Struct truncate to scalar
// CHECK-LABEL: call9
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[D2:%.*]] = alloca double, align 8
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
@@ -204,6 +205,7 @@ export void call10(int4 I) {
// truncate derived struct
// CHECK-LABEL: call11
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[B:%.*]] = alloca %struct.BFields, align 1
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl
index 5c6940c4d2e4d..023c369059bcb 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/StructPassing.hlsl
@@ -13,6 +13,7 @@ struct S : P {
// CHECK-LABEL: define hidden void @_Z5case11S(
// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[SLOCAL:%.*]] = alloca [[STRUCT_S:%.*]], align 1
// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
@@ -27,6 +28,7 @@ void case1(S s) {
// CHECK-LABEL: define hidden void @_Z5case21S(
// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[SLOCAL:%.*]] = alloca [[STRUCT_S:%.*]], align 1
// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_S]], align 1
@@ -46,6 +48,7 @@ void useS(S s) {}
// CHECK-LABEL: define hidden void @_Z5case31S(
// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[BYVAL_TEMP:%.*]] = alloca [[STRUCT_S:%.*]], align 1
// CHECK-NEXT: store ptr [[S]], ptr [[S_INDIRECT_ADDR]], align 4
@@ -61,6 +64,7 @@ void case3(S s) {
// CHECK-LABEL: define hidden void @_Z5case41S(
// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[PLOCAL:%.*]] = alloca [[STRUCT_P:%.*]], align 1
// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 1
@@ -77,6 +81,7 @@ void case4(S s) {
// CHECK-LABEL: define hidden void @_Z5case51S(
// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[P:%.*]] = alloca [[STRUCT_P:%.*]], align 1
// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_P]], align 1
@@ -98,6 +103,7 @@ void useP(P p) {}
// CHECK-LABEL: define hidden void @_Z5case61S(
// CHECK-SAME: ptr noundef dead_on_return [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[AGG_TMP:%.*]] = alloca [[STRUCT_P:%.*]], align 1
// CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 1
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl
index 31263bca443a6..89f5ff8f16825 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl
@@ -95,6 +95,7 @@ struct Derived : BFields {
// vector flat cast from derived struct with bitfield
// CHECK-LABEL: call6
// CHECK-NEXT: entry:
+// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[D_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[A:%.*]] = alloca <4 x i32>, align 4
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
@@ -221,6 +222,7 @@ struct BoolVecStruct {
// vector flat cast from struct containing bool vector
// CHECK-LABEL: call10
// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
// CHECK-NEXT: [[S_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[V:%.*]] = alloca <2 x i32>, align 4
// CHECK-NEXT: [[AGG_TEMP:%.*]] = alloca %struct.BoolVecStruct, align 1
diff --git a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
index d1eb2f740dbf2..d29ec8e9f87b9 100644
--- a/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
+++ b/clang/test/CodeGenHLSL/resources/cbuffer_struct_passing.hlsl
@@ -2,7 +2,7 @@
// RUN: FileCheck %s -DCONST_ADDR_SPACE=2 -DPADDING_TYPE="dx.Padding"
// RUN: %clang_cc1 -triple spirv-pc-vulkan1.3-library -finclude-default-header -emit-llvm -disable-llvm-passes -o - %s | \
-// RUN: FileCheck %s -DCONST_ADDR_SPACE=12 -DPADDING_TYPE="spirv.Padding" --check-prefixes=CHECK,SPIRV
+// RUN: FileCheck %s -DCONST_ADDR_SPACE=12 -DPADDING_TYPE="spirv.Padding"
struct P {
float3 a;
@@ -33,7 +33,7 @@ cbuffer CB {
// CHECK-LABEL: case1
// CHECK-NEXT: entry:
-// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
//
// Copy S field by field into local variable in default address space.
//
@@ -57,7 +57,7 @@ void case1() {
// CHECK-LABEL: case2
// CHECK-NEXT: entry:
-// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
//
// Copy S field by field into a temporary variable in default address space.
//
@@ -130,7 +130,7 @@ void case2() {
// CHECK-LABEL: case3
// CHECK-NEXT: entry:
-// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
void case3() {
// CHECK-NEXT: [[LocalT:%.*]] = alloca %struct.T, align 1
@@ -191,7 +191,7 @@ void case3() {
// CHECK-LABEL: case4
// CHECK-NEXT: entry:
-// SPIRV-NEXT: call token @llvm.experimental.convergence.entry()
+// CHECK-NEXT: call token @llvm.experimental.convergence.entry()
void case4() {
// CHECK-NEXT: [[LocalP1:%.*]] = alloca %struct.P, align 1
diff --git a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
index 34c57ed76a67e..8ebc5f696e33e 100644
--- a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
+++ b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
@@ -37,8 +37,8 @@ void main() {
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[Another]], i32 8, i1 false)
// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
-// CHECK-NEXT: %0 = load i32, ptr [[First]], align 1
-// CHECK-NEXT: ret i32 %0
+// CHECK-NEXT: %[[LOAD:.*]] = load i32, ptr [[First]], align 1
+// CHECK-NEXT: ret i32 %[[LOAD]]
// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair9getSecondEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
// CHECK-NEXT:entry:
@@ -53,5 +53,5 @@ void main() {
// CHECK-NEXT: store i32 123, ptr [[Second]], align 1
// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
// CHECK-NEXT: [[Second2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
-// CHECK-NEXT: %0 = load i32, ptr [[Second2]], align 1
-// CHECK-NEXT: ret i32 %0
+// CHECK-NEXT: %[[LOAD:.*]] = load i32, ptr [[Second2]], align 1
+// CHECK-NEXT: ret i32 %[[LOAD]]
>From 10ad30d343ae340dbc61a4f2f8ea649f8d71d7a0 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Wed, 3 Jun 2026 12:52:12 -0700
Subject: [PATCH 16/22] remove stray files
---
clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak | 160 ------------------
.../this-assignment-overload.hlsl.bak | 58 -------
2 files changed, 218 deletions(-)
delete mode 100644 clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak
delete mode 100644 clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak
diff --git a/clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak b/clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak
deleted file mode 100644
index a17e0ae6038b1..0000000000000
--- a/clang/test/CodeGenHLSL/BoolMatrix.hlsl.bak
+++ /dev/null
@@ -1,160 +0,0 @@
-// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
-// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
-
-// CHECK: [[ConstInit_fn3_s:@.*]] = private unnamed_addr constant { <4 x i1>, [12 x i8], float } { <4 x i1> <i1 true, i1 false, i1 true, i1 false>, [12 x i8] undef, float 1.000000e+00 }, align 1
-// CHECK: [[ConstInit_fn4_Arr:@.*]] = private unnamed_addr constant [2 x <4 x i1>] [<4 x i1> splat (i1 true), <4 x i1> zeroinitializer], align 4
-// CHECK: [[ConstInit_fn6_s:@.*]] = private unnamed_addr constant { <4 x i1>, [12 x i8], float } { <4 x i1> <i1 true, i1 false, i1 true, i1 false>, [12 x i8] undef, float 1.000000e+00 }, align 1
-// CHECK: [[ConstInit_fn7_Arr:@.*]] = private unnamed_addr constant [2 x <4 x i1>] [<4 x i1> splat (i1 true), <4 x i1> zeroinitializer], align 4
-
-
-struct S {
- bool2x2 bM;
- float f;
-};
-
-// CHECK-LABEL: define hidden noundef i1 @_Z3fn1v(
-// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[RETVAL:%.*]] = alloca i1, align 4
-// CHECK-NEXT: [[B:%.*]] = alloca [2 x <2 x i32>], align 4
-// CHECK-NEXT: store <4 x i32> splat (i32 1), ptr [[B]], align 4
-// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[B]], align 4
-// CHECK-NEXT: [[MATRIXEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 0
-// CHECK-NEXT: store i32 [[MATRIXEXT]], ptr [[RETVAL]], align 4
-// CHECK-NEXT: [[TMP1:%.*]] = load i1, ptr [[RETVAL]], align 4
-// CHECK-NEXT: ret i1 [[TMP1]]
-//
-bool fn1() {
- bool2x2 B = {true,true,true,true};
- return B[0][0];
-}
-
-// CHECK-LABEL: define hidden noundef <4 x i1> @_Z3fn2b(
-// CHECK-SAME: i1 noundef [[V:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x i1>, align 4
-// CHECK-NEXT: [[V_ADDR:%.*]] = alloca i32, align 4
-// CHECK-NEXT: [[A:%.*]] = alloca [2 x <2 x i32>], align 4
-// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[V]] to i32
-// CHECK-NEXT: store i32 [[STOREDV]], ptr [[V_ADDR]], align 4
-// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[V_ADDR]], align 4
-// CHECK-NEXT: [[LOADEDV:%.*]] = icmp ne i32 [[TMP0]], 0
-// CHECK-NEXT: [[VECINIT:%.*]] = insertelement <4 x i1> poison, i1 [[LOADEDV]], i32 0
-// CHECK-NEXT: [[VECINIT1:%.*]] = insertelement <4 x i1> [[VECINIT]], i1 true, i32 2
-// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[V_ADDR]], align 4
-// CHECK-NEXT: [[LOADEDV2:%.*]] = icmp ne i32 [[TMP1]], 0
-// CHECK-NEXT: [[VECINIT3:%.*]] = insertelement <4 x i1> [[VECINIT1]], i1 [[LOADEDV2]], i32 1
-// CHECK-NEXT: [[VECINIT4:%.*]] = insertelement <4 x i1> [[VECINIT3]], i1 false, i32 3
-// CHECK-NEXT: [[TMP2:%.*]] = zext <4 x i1> [[VECINIT4]] to <4 x i32>
-// CHECK-NEXT: store <4 x i32> [[TMP2]], ptr [[A]], align 4
-// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[A]], align 4
-// CHECK-NEXT: store <4 x i32> [[TMP3]], ptr [[RETVAL]], align 4
-// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i1>, ptr [[RETVAL]], align 4
-// CHECK-NEXT: ret <4 x i1> [[TMP4]]
-//
-bool2x2 fn2(bool V) {
- bool2x2 A = {V, true, V, false};
- return A;
-}
-
-// CHECK-LABEL: define hidden noundef i1 @_Z3fn3v(
-// CHECK-SAME: ) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[RETVAL:%.*]] = alloca i1, align 4
-// CHECK-NEXT: [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 1
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[S]], ptr align 1 [[ConstInit_fn3_s]], i32 20, i1 false)
-// CHECK-NEXT: [[BM:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[S]], i32 0, i32 0
-// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[BM]], align 1
-// CHECK-NEXT: [[MATRIXEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 0
-// CHECK-NEXT: store i32 [[MATRIXEXT]], ptr [[RETVAL]], align 4
-// CHECK-NEXT: [[TMP1:%.*]] = load i1, ptr [[RETVAL]], align 4
-// CHECK-NEXT: ret i1 [[TMP1]]
-//
-bool fn3() {
- S s = {{true,true,false,false}, 1.0};
- return s.bM[0][0];
-}
-
-// CHECK-LABEL: define hidden noundef i1 @_Z3fn4v(
-// CHECK-SAME: ) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[RETVAL:%.*]] = alloca i1, align 4
-// CHECK-NEXT: [[ARR:%.*]] = alloca [2 x [2 x <2 x i32>]], align 4
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 %Arr, ptr align 4 [[ConstInit_fn4_Arr]], i32 8, i1 false)
-// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x [2 x <2 x i32>]], ptr [[ARR]], i32 0, i32 0
-// CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[ARRAYIDX]], align 4
-// CHECK-NEXT: [[MATRIXEXT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 1
-// CHECK-NEXT: store i32 [[MATRIXEXT]], ptr [[RETVAL]], align 4
-// CHECK-NEXT: [[TMP1:%.*]] = load i1, ptr [[RETVAL]], align 4
-// CHECK-NEXT: ret i1 [[TMP1]]
-//
-bool fn4() {
- bool2x2 Arr[2] = {{true,true,true,true}, {false,false,false,false}};
- return Arr[0][1][0];
-}
-
-// CHECK-LABEL: define hidden void @_Z3fn5v(
-// CHECK-SAME: ) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[M:%.*]] = alloca [2 x <2 x i32>], align 4
-// CHECK-NEXT: store <4 x i32> splat (i32 1), ptr [[M]], align 4
-// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <4 x i32>, ptr [[M]], i32 0, i32 3
-// CHECK-NEXT: store i32 0, ptr [[TMP0]], align 4
-// CHECK-NEXT: ret void
-//
-void fn5() {
- bool2x2 M = {true,true,true,true};
- M[1][1] = false;
-}
-
-// CHECK-LABEL: define hidden void @_Z3fn6v(
-// CHECK-SAME: ) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[V:%.*]] = alloca i32, align 4
-// CHECK-NEXT: [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 1
-// CHECK-NEXT: store i32 0, ptr [[V]], align 4
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[S]], ptr align 1 [[ConstInit_fn6_s]], i32 20, i1 false)
-// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[V]], align 4
-// CHECK-NEXT: [[LOADEDV:%.*]] = icmp ne i32 [[TMP0]], 0
-// CHECK-NEXT: [[BM:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[S]], i32 0, i32 0
-// CHECK-NEXT: [[TMP1:%.*]] = zext i1 [[LOADEDV]] to i32
-// CHECK-NEXT: [[TMP2:%.*]] = getelementptr <4 x i32>, ptr [[BM]], i32 0, i32 1
-// CHECK-NEXT: store i32 [[TMP1]], ptr [[TMP2]], align 4
-// CHECK-NEXT: ret void
-//
-void fn6() {
- bool V = false;
- S s = {{true,true,false,false}, 1.0};
- s.bM[1][0] = V;
-}
-
-// CHECK-LABEL: define hidden void @_Z3fn7v(
-// CHECK-SAME: ) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[ARR:%.*]] = alloca [2 x [2 x <2 x i32>]], align 4
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARR]], ptr align 4 [[ConstInit_fn7_Arr]], i32 8, i1 false)
-// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x [2 x <2 x i32>]], ptr [[ARR]], i32 0, i32 0
-// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <4 x i32>, ptr [[ARRAYIDX]], i32 0, i32 1
-// CHECK-NEXT: store i32 0, ptr [[TMP0]], align 4
-// CHECK-NEXT: ret void
-//
-void fn7() {
- bool2x2 Arr[2] = {{true,true,true,true}, {false,false,false,false}};
- Arr[0][1][0] = false;
-}
-
-// CHECK-LABEL: define hidden noundef <16 x i1> @_Z3fn8u11matrix_typeILm4ELm4EbE(
-// CHECK-SAME: <16 x i1> noundef [[M:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT: [[ENTRY:.*:]]
-// CHECK-NEXT: [[RETVAL:%.*]] = alloca <16 x i1>, align 4
-// CHECK-NEXT: [[M_ADDR:%.*]] = alloca [4 x <4 x i32>], align 4
-// CHECK-NEXT: [[TMP0:%.*]] = zext <16 x i1> [[M]] to <16 x i32>
-// CHECK-NEXT: store <16 x i32> [[TMP0]], ptr [[M_ADDR]], align 4
-// CHECK-NEXT: [[TMP1:%.*]] = load <16 x i32>, ptr [[M_ADDR]], align 4
-// CHECK-NEXT: store <16 x i32> [[TMP1]], ptr [[RETVAL]], align 4
-// CHECK-NEXT: [[TMP2:%.*]] = load <16 x i1>, ptr [[RETVAL]], align 4
-// CHECK-NEXT: ret <16 x i1> [[TMP2]]
-//
-bool4x4 fn8(bool4x4 m) {
- return m;
-}
diff --git a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak
deleted file mode 100644
index 7b08f7b96c624..0000000000000
--- a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl.bak
+++ /dev/null
@@ -1,58 +0,0 @@
-// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -disable-llvm-passes -o - -std=hlsl202x %s | FileCheck %s
-
-struct Pair {
- int First;
- int Second;
- int getFirst() {
- Pair Another = {5, 10};
- this = Another;
- return this.First;
- }
- int getSecond() {
- this = {0, 123};
- return Second;
- }
- void operator=(Pair P) {
- First = P.First;
- Second = 2;
- }
-};
-[numthreads(1, 1, 1)]
-void main() {
- Pair Vals = {1, 2};
- Vals.First = Vals.getFirst();
- Vals.Second = Vals.getSecond();
-}
-
-// This test makes a probably safe assumption that HLSL 202x includes operator overloading for assignment operators.
-// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair8getFirstEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
-// CHECK-NEXT: %Another = alloca %struct.Pair, align 1
-// CHECK-NEXT: [[AggTmp:%.*]] = alloca %struct.Pair, align 1
-// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
-// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
-// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr %Another, i32 0, i32 0
-// CHECK-NEXT: store i32 5, ptr [[First]], align 1
-// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr %Another, i32 0, i32 1
-// CHECK-NEXT: store i32 10, ptr [[Second]], align 1
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[AggTmp]], ptr align 1 %Another, i32 8, i1 false)
-// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[AggTmp]])
-// CHECK-NEXT: [[First2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
-// CHECK-NEXT: %0 = load i32, ptr [[First2]], align 1
-// CHECK-NEXT: ret i32 %0
-
-// CHECK: define linkonce_odr hidden noundef i32 @_ZN4Pair9getSecondEv(ptr noundef nonnull align 1 dereferenceable(8) %this) #0 align 2 {
-// CHECK-NEXT:entry:
-// CHECK-NEXT: [[ThisPtrAdds:%.*]] = alloca ptr, align 4
-// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
-// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
-// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
-// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 0
-// CHECK-NEXT: store i32 0, ptr [[First]], align 1
-// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 1
-// CHECK-NEXT: store i32 123, ptr [[Second]], align 1
-// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
-// CHECK-NEXT: [[Second2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
-// CHECK-NEXT: %0 = load i32, ptr [[Second2]], align 1
-// CHECK-NEXT: ret i32 %0
>From e905a3e04d18b72c12a47dbdfbe25acb827ec50c Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 4 Jun 2026 09:54:39 -0700
Subject: [PATCH 17/22] Add isHLSLBuiltinRecord bit to CXXRecordDecl definiton
data and use it instead of hasUserProvidedSpecialMembers()
---
.../clang/AST/CXXRecordDeclDefinitionBits.def | 4 +
clang/include/clang/AST/DeclCXX.h | 80 ++++++-------------
clang/lib/AST/Type.cpp | 3 +-
clang/lib/Sema/HLSLBuiltinTypeDeclBuilder.cpp | 1 +
clang/lib/Sema/SemaHLSL.cpp | 6 +-
clang/lib/Sema/SemaInit.cpp | 4 +-
clang/lib/Sema/SemaOverload.cpp | 2 +-
clang/lib/Sema/SemaTemplateInstantiate.cpp | 2 +
.../CodeGenHLSL/this-assignment-overload.hlsl | 10 +--
9 files changed, 44 insertions(+), 68 deletions(-)
diff --git a/clang/include/clang/AST/CXXRecordDeclDefinitionBits.def b/clang/include/clang/AST/CXXRecordDeclDefinitionBits.def
index 7e6e2147a448d..97e61aaec7d51 100644
--- a/clang/include/clang/AST/CXXRecordDeclDefinitionBits.def
+++ b/clang/include/clang/AST/CXXRecordDeclDefinitionBits.def
@@ -253,6 +253,10 @@ FIELD(IsAnyDestructorNoReturn, 1, NO_MERGE)
/// type that is intangible). HLSL only.
FIELD(IsHLSLIntangible, 1, NO_MERGE)
+/// Whether the record type is a built-in HLSL record which must be handled
+/// differently by the compiler than user-defined records. HLSL only.
+FIELD(IsHLSLBuiltinRecord, 1, NO_MERGE)
+
/// Whether the pointer fields in this class should have pointer field
/// protection (PFP) by default, either because of an attribute, the
/// -fexperimental-pointer-field-protection-abi compiler flag or inheritance
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index 36cd8f519ea34..868a9bb52215c 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -759,38 +759,20 @@ class CXXRecordDecl : public RecordDecl {
needsImplicitDefaultConstructor();
}
- // Used by HLSL to determine if a record is a built-in implicit HLSL
- // struct/class or a user-defined one. User-defined HLSL records cannot
- // have ctors, dtors, or overloaded operators, while implicit built-in
- // HLSL records such as resource classes can. It would be nice to use the
- // isImplicit() methods to determine that, but this flag is not propagated
- // to template-instantiated classes.
- //
- /// Determines whether this class has any user provided special members.
- bool hasUserProvidedSpecialMembers() const {
- return data().UserDeclaredSpecialMembers & SMF_All ||
- data().UserDeclaredConstructor ||
- data().UserProvidedDefaultConstructor;
- }
-
/// Determine if we need to declare a default constructor for
/// this class.
///
/// This value is used for lazy creation of default constructors.
bool needsImplicitDefaultConstructor() const {
- // In HLSL, only built-in records like resource classes can have
- // constructors and overloadable operators.
- if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
- return false;
-
- return (!data().UserDeclaredConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
- // FIXME: Proposed fix to core wording issue: if a class inherits
- // a default constructor and doesn't explicitly declare one, one
- // is declared implicitly.
- (data().HasInheritedDefaultConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor));
+ return ((!data().UserDeclaredConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
+ // FIXME: Proposed fix to core wording issue: if a class inherits
+ // a default constructor and doesn't explicitly declare one, one
+ // is declared implicitly.
+ (data().HasInheritedDefaultConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor))) &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord());
}
/// Determine whether this class has any user-declared constructors.
@@ -816,12 +798,8 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// constructor to be lazily declared.
bool needsImplicitCopyConstructor() const {
- // In HLSL, only built-in records like resource classes can have
- // constructors and overloadable operators.
- if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
- return false;
-
- return !(data().DeclaredSpecialMembers & SMF_CopyConstructor);
+ return !(data().DeclaredSpecialMembers & SMF_CopyConstructor) &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord());
}
/// Determine whether we need to eagerly declare a defaulted copy
@@ -914,16 +892,11 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class should get an implicit move
/// constructor or if any existing special member function inhibits this.
bool needsImplicitMoveConstructor() const {
- // In HLSL, only built-in records like resource classes can have
- // constructors and overloadable operators.
- if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
- return false;
-
return !(data().DeclaredSpecialMembers & SMF_MoveConstructor) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveAssignment() &&
- !hasUserDeclaredDestructor();
+ !hasUserDeclaredMoveAssignment() && !hasUserDeclaredDestructor() &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord());
}
/// Determine whether we need to eagerly declare a defaulted move
@@ -952,12 +925,8 @@ class CXXRecordDecl : public RecordDecl {
/// Determine whether this class needs an implicit copy
/// assignment operator to be lazily declared.
bool needsImplicitCopyAssignment() const {
- // In HLSL, only built-in records like resource classes can have
- // constructors and overloadable operators.
- if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
- return false;
-
- return !(data().DeclaredSpecialMembers & SMF_CopyAssignment);
+ return !(data().DeclaredSpecialMembers & SMF_CopyAssignment) &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord());
}
/// Determine whether we need to eagerly declare a defaulted copy
@@ -1015,17 +984,12 @@ class CXXRecordDecl : public RecordDecl {
/// assignment operator or if any existing special member function inhibits
/// this.
bool needsImplicitMoveAssignment() const {
- // In HLSL, only built-in records like resource classes can have
- // constructors and overloadable operators.
- if (getLangOpts().HLSL && !hasUserProvidedSpecialMembers())
- return false;
-
return !(data().DeclaredSpecialMembers & SMF_MoveAssignment) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveConstructor() &&
- !hasUserDeclaredDestructor() &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable());
+ !hasUserDeclaredMoveConstructor() && !hasUserDeclaredDestructor() &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable()) &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord());
}
/// Determine whether we need to eagerly declare a move assignment
@@ -1594,6 +1558,14 @@ class CXXRecordDecl : public RecordDecl {
/// a field or in base class.
bool isHLSLIntangible() const { return data().IsHLSLIntangible; }
+ /// Returns true if the class is a built-in HLSL record.
+ bool isHLSLBuiltinRecord() const { return data().IsHLSLBuiltinRecord; }
+
+ /// Sets the flag that the class is a built-in HLSL record.
+ void setIsHLSLBuiltinRecord(bool Value) {
+ data().IsHLSLBuiltinRecord = Value;
+ }
+
/// If the class is a local class [class.local], returns
/// the enclosing function declaration.
const FunctionDecl *isLocalClass() const {
diff --git a/clang/lib/AST/Type.cpp b/clang/lib/AST/Type.cpp
index d4504e6b04f8c..43b8cdc686cb9 100644
--- a/clang/lib/AST/Type.cpp
+++ b/clang/lib/AST/Type.cpp
@@ -3303,8 +3303,7 @@ bool QualType::isHLSLPODType(const ASTContext &Context) const {
dyn_cast_or_null<CXXRecordDecl>(BaseTy->getAsRecordDecl())) {
// User-defined records in HLSL do not have constructors or copy/assignment
// operators. They are still considered POD.
- if (!RD->hasUserProvidedSpecialMembers() || RD->isTrivial() ||
- RD->isStandardLayout())
+ if (!RD->isHLSLBuiltinRecord() || RD->isTrivial() || RD->isStandardLayout())
return true;
}
return false;
diff --git a/clang/lib/Sema/HLSLBuiltinTypeDeclBuilder.cpp b/clang/lib/Sema/HLSLBuiltinTypeDeclBuilder.cpp
index de170c86400d2..40e4f4024d6da 100644
--- a/clang/lib/Sema/HLSLBuiltinTypeDeclBuilder.cpp
+++ b/clang/lib/Sema/HLSLBuiltinTypeDeclBuilder.cpp
@@ -2167,6 +2167,7 @@ BuiltinTypeDeclBuilder &BuiltinTypeDeclBuilder::completeDefinition() {
"Definition must be started before completing it.");
Record->completeDefinition();
+ Record->setIsHLSLBuiltinRecord(true);
return *this;
}
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index c201f24cf5049..24409088b62d8 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -5650,7 +5650,7 @@ bool SemaHLSL::ActOnUninitializedVarDecl(VarDecl *VD) {
while (Ty->isArrayType())
Ty = Ty->getArrayElementTypeNoTypeQual()->getUnqualifiedDesugaredType();
if (CXXRecordDecl *RD = Ty->getAsCXXRecordDecl())
- return !RD->hasUserProvidedSpecialMembers();
+ return !RD->isHLSLBuiltinRecord();
return false;
}
@@ -5751,9 +5751,7 @@ bool SemaHLSL::canHaveOverloadedBinOp(QualType LHSTy, BinaryOperatorKind Opc) {
CXXRecordDecl *RD = LHSTy->getAsCXXRecordDecl();
if (!RD)
return true;
- // hasUserProvidedSpecialMembers() should be true only for HLSL built-in
- // records like resources.
- return RD->hasUserProvidedSpecialMembers() || Opc != BO_Assign;
+ return RD->isHLSLBuiltinRecord() || Opc != BO_Assign;
}
// Walks though the global variable declaration, collects all resource binding
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index 745f1aef40ea9..8f685feac4beb 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -6937,7 +6937,7 @@ void InitializationSequence::InitializeFrom(Sema &S,
// constructors or conversion functions)
if (DestType->isRecordType() &&
(!S.getLangOpts().HLSL ||
- DestType->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers())) {
+ DestType->getAsCXXRecordDecl()->isHLSLBuiltinRecord())) {
// - If the initialization is direct-initialization, or if it is
// copy-initialization where the cv-unqualified version of the
// source type is the same class as, or a derived class of, the
@@ -7026,7 +7026,7 @@ void InitializationSequence::InitializeFrom(Sema &S,
// constructors or conversion functions).
if (!SourceType.isNull() && SourceType->isRecordType() &&
(!S.getLangOpts().HLSL ||
- SourceType->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers())) {
+ SourceType->getAsCXXRecordDecl()->isHLSLBuiltinRecord())) {
assert(Initializer && "Initializer must be non-null");
// For a conversion to _Atomic(T) from either T or a class type derived
// from T, initialize the T object then convert to _Atomic type.
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 70ad3adc636e3..b3e0d5a48a521 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -15537,7 +15537,7 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
if (Opc == BO_Assign &&
(!LHSTy->isOverloadableType() ||
(getLangOpts().HLSL && LHSTy->isRecordType() &&
- !LHSTy->getAsCXXRecordDecl()->hasUserProvidedSpecialMembers())))
+ !LHSTy->getAsCXXRecordDecl()->isHLSLBuiltinRecord())))
return CreateBuiltinBinOp(OpLoc, Opc, Args[0], Args[1]);
// Build the overload set.
diff --git a/clang/lib/Sema/SemaTemplateInstantiate.cpp b/clang/lib/Sema/SemaTemplateInstantiate.cpp
index f168c99d1ac1a..758abc2c2bac1 100644
--- a/clang/lib/Sema/SemaTemplateInstantiate.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiate.cpp
@@ -3727,6 +3727,8 @@ bool Sema::InstantiateClassImpl(
}
}
+ Instantiation->setIsHLSLBuiltinRecord(Pattern->isHLSLBuiltinRecord());
+
// Exit the scope of this instantiation.
SavedContext.pop();
diff --git a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
index 8ebc5f696e33e..bf5a125d345d1 100644
--- a/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
+++ b/clang/test/CodeGenHLSL/this-assignment-overload.hlsl
@@ -34,8 +34,8 @@ void main() {
// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Another]], ptr align 1 @__const._ZN4Pair8getFirstEv.Another, i32 8, i1 false)
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[Another]], i32 8, i1 false)
-// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[ThisPtr]], ptr align 1 [[Another]], i32 8, i1 false)
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[ThisPtr]], i32 8, i1 false)
// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
// CHECK-NEXT: %[[LOAD:.*]] = load i32, ptr [[First]], align 1
// CHECK-NEXT: ret i32 %[[LOAD]]
@@ -47,11 +47,11 @@ void main() {
// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Pair, align 1
// CHECK-NEXT: store ptr %this, ptr [[ThisPtrAdds]], align 4
// CHECK-NEXT: [[ThisPtr:%.*]] = load ptr, ptr [[ThisPtrAdds]], align 4
-// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 0
+// CHECK-NEXT: [[First:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 0
// CHECK-NEXT: store i32 0, ptr [[First]], align 1
-// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[Tmp]], i32 0, i32 1
+// CHECK-NEXT: [[Second:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
// CHECK-NEXT: store i32 123, ptr [[Second]], align 1
-// CHECK-NEXT: call void @_ZN4PairaSES_(ptr noundef nonnull align 1 dereferenceable(8) [[ThisPtr]], ptr noundef byval(%struct.Pair) align 1 [[Tmp]])
+// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[ThisPtr]], i32 8, i1 false)
// CHECK-NEXT: [[Second2:%.*]] = getelementptr inbounds nuw %struct.Pair, ptr [[ThisPtr]], i32 0, i32 1
// CHECK-NEXT: %[[LOAD:.*]] = load i32, ptr [[Second2]], align 1
// CHECK-NEXT: ret i32 %[[LOAD]]
>From fa04e38add72570dc8f33e34452ef0435920bb64 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 4 Jun 2026 14:53:06 -0700
Subject: [PATCH 18/22] code review feedback
---
clang/include/clang/AST/TypeBase.h | 11 +++----
clang/lib/AST/Type.cpp | 31 ++++++++-----------
clang/lib/Sema/SemaExpr.cpp | 15 +++++----
clang/lib/Sema/SemaOverload.cpp | 5 ++-
.../BasicFeatures/ArrayElementwiseCast.hlsl | 10 ++----
5 files changed, 31 insertions(+), 41 deletions(-)
diff --git a/clang/include/clang/AST/TypeBase.h b/clang/include/clang/AST/TypeBase.h
index 779a4eba3468d..3a9918b33082a 100644
--- a/clang/include/clang/AST/TypeBase.h
+++ b/clang/include/clang/AST/TypeBase.h
@@ -1111,11 +1111,6 @@ class QualType {
/// CXXRecordDecl::isCXX11StandardLayout, this takes DRs into account.
bool isCXX11PODType(const ASTContext &Context) const;
- /// Return true if this is a POD type according to the HLSL rules.
- /// User-defined records in HLSL do not have default constructors, which
- /// is a POD requirement for C++.
- bool isHLSLPODType(const ASTContext &Context) const;
-
/// Return true if this is a trivial type per (C++0x [basic.types]p9)
bool isTrivialType(const ASTContext &Context) const;
@@ -2803,8 +2798,10 @@ class alignas(TypeAlignment) Type : public ExtQualsTypeCommonBase {
bool isHLSLInlineSpirvType() const;
bool isHLSLResourceRecord() const;
bool isHLSLResourceRecordArray() const;
- bool isHLSLIntangibleType()
- const; // Any HLSL intangible type (builtin, array, class)
+ // Any HLSL intangible type (builtin, array, class)
+ bool isHLSLIntangibleType() const;
+ // User-defined HLSL records or arrays of such records
+ bool isHLSLStandardRecordOrArrayOf() const;
/// Determines if this type, which must satisfy
/// isObjCLifetimeType(), is implicitly __unsafe_unretained rather
diff --git a/clang/lib/AST/Type.cpp b/clang/lib/AST/Type.cpp
index 43b8cdc686cb9..cbead120cbffc 100644
--- a/clang/lib/AST/Type.cpp
+++ b/clang/lib/AST/Type.cpp
@@ -2786,9 +2786,9 @@ QualType Type::getRVVEltType(const ASTContext &Ctx) const {
}
bool QualType::isPODType(const ASTContext &Context) const {
- // HLSL has a more relaxed definition of POD than C++11.
- if (Context.getLangOpts().HLSL)
- return isHLSLPODType(Context);
+ if (Context.getLangOpts().HLSL &&
+ getTypePtr()->isHLSLStandardRecordOrArrayOf())
+ return true;
// C++11 has a more relaxed definition of POD.
if (Context.getLangOpts().CPlusPlus11)
@@ -3294,21 +3294,6 @@ bool QualType::isCXX11PODType(const ASTContext &Context) const {
return false;
}
-bool QualType::isHLSLPODType(const ASTContext &Context) const {
- if (isCXX11PODType(Context))
- return true;
-
- const Type *BaseTy = getTypePtr()->getBaseElementTypeUnsafe();
- if (const auto *RD =
- dyn_cast_or_null<CXXRecordDecl>(BaseTy->getAsRecordDecl())) {
- // User-defined records in HLSL do not have constructors or copy/assignment
- // operators. They are still considered POD.
- if (!RD->isHLSLBuiltinRecord() || RD->isTrivial() || RD->isStandardLayout())
- return true;
- }
- return false;
-}
-
bool Type::isNothrowT() const {
if (const auto *RD = getAsCXXRecordDecl()) {
IdentifierInfo *II = RD->getIdentifier();
@@ -5548,6 +5533,16 @@ bool Type::isHLSLIntangibleType() const {
return RD->isHLSLIntangible();
}
+bool Type::isHLSLStandardRecordOrArrayOf() const {
+ const Type *BaseTy = getBaseElementTypeUnsafe();
+ if (const auto *RD =
+ dyn_cast_or_null<CXXRecordDecl>(BaseTy->getAsRecordDecl())) {
+ if (!RD->isHLSLBuiltinRecord())
+ return true;
+ }
+ return false;
+}
+
QualType::DestructionKind QualType::isDestructedTypeImpl(QualType type) {
switch (type.getObjCLifetime()) {
case Qualifiers::OCL_None:
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index fbd4d275f5f7f..aa32332dd577d 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -16117,12 +16117,15 @@ ExprResult Sema::BuildBinOp(Scope *S, SourceLocation OpLoc,
}
if (getLangOpts().CPlusPlus) {
- if ((LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent() ||
- LHSExpr->getType()->isOverloadableType() ||
- RHSExpr->getType()->isOverloadableType()) &&
- (!getLangOpts().HLSL ||
- HLSL().canHaveOverloadedBinOp(LHSExpr->getType(), Opc) ||
- HLSL().canHaveOverloadedBinOp(RHSExpr->getType(), Opc)))
+ bool CanOverloadBinOp =
+ !getLangOpts().HLSL ||
+ HLSL().canHaveOverloadedBinOp(LHSExpr->getType(), Opc) ||
+ HLSL().canHaveOverloadedBinOp(RHSExpr->getType(), Opc);
+ bool TypeDependent =
+ LHSExpr->isTypeDependent() || RHSExpr->isTypeDependent();
+ bool Overloadable = LHSExpr->getType()->isOverloadableType() ||
+ RHSExpr->getType()->isOverloadableType();
+ if (CanOverloadBinOp && (TypeDependent || Overloadable))
return BuildOverloadedBinOp(*this, S, OpLoc, Opc, LHSExpr, RHSExpr);
}
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index b3e0d5a48a521..543c97981bd6a 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -1831,9 +1831,6 @@ TryImplicitConversion(Sema &S, Expr *From, QualType ToType,
// given Conversion rank, in spite of the fact that a copy/move
// constructor (i.e., a user-defined conversion function) is
// called for those cases.
- // HLSL:
- // A conversion of an expression of class type to the same class
- // type needs implicit LvaluetoRvalue conversion.
QualType FromType = From->getType();
if (ToType->isRecordType() &&
(S.Context.hasSameUnqualifiedType(FromType, ToType) ||
@@ -1849,6 +1846,8 @@ TryImplicitConversion(Sema &S, Expr *From, QualType ToType,
// appropriate constructor to copy the returned object, if needed.
ICS.Standard.CopyConstructor = nullptr;
+ // In HLSL, a conversion of an expression of class type to the same class
+ // type needs implicit LvaluetoRvalue conversion.
if (S.getLangOpts().HLSL)
ICS.Standard.First = ICK_Lvalue_To_Rvalue;
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
index b8611914efb29..f684eecf20ff1 100644
--- a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
+++ b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl
@@ -156,13 +156,9 @@ struct Derived : BFields {
// flatten from a derived struct with bitfields
// CHECK-LABEL: call8
-// CHECK-NEXT: entry:
-// CHECK-NEXT: %[[#C_ENTRY:]] = call token @llvm.experimental.convergence.entry()
-// CHECK-NEXT: [[DIndirectAddr:%.*]] = alloca ptr, align 4
-// CHECK-NEXT: [[A:%.*]] = alloca [4 x i32], align 4
-// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1
-// CHECK-NEXT: store ptr %D, ptr [[DIndirectAddr]], align 4
-// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
+// CHECK: [[A:%.*]] = alloca [4 x i32], align 4
+// CHECK: [[Tmp:%.*]] = alloca %struct.Derived, align 1
+// CHECK: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false)
// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 0
// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 1
// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 2
>From 38926b979fc7381744cbc2726a3ee64ebdfa9951 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Mon, 8 Jun 2026 17:57:43 -0700
Subject: [PATCH 19/22] after-merge fixes and function rename
---
clang/include/clang/AST/TypeBase.h | 4 ++--
clang/lib/AST/Type.cpp | 6 +++---
clang/lib/CodeGen/CGExprAgg.cpp | 6 +++++-
3 files changed, 10 insertions(+), 6 deletions(-)
diff --git a/clang/include/clang/AST/TypeBase.h b/clang/include/clang/AST/TypeBase.h
index 3a9918b33082a..65a35342fb09b 100644
--- a/clang/include/clang/AST/TypeBase.h
+++ b/clang/include/clang/AST/TypeBase.h
@@ -2800,8 +2800,8 @@ class alignas(TypeAlignment) Type : public ExtQualsTypeCommonBase {
bool isHLSLResourceRecordArray() const;
// Any HLSL intangible type (builtin, array, class)
bool isHLSLIntangibleType() const;
- // User-defined HLSL records or arrays of such records
- bool isHLSLStandardRecordOrArrayOf() const;
+ // User-defined HLSL records or arrays of such records in standard layout
+ bool isHLSLStandardLayoutRecordOrArrayOf() const;
/// Determines if this type, which must satisfy
/// isObjCLifetimeType(), is implicitly __unsafe_unretained rather
diff --git a/clang/lib/AST/Type.cpp b/clang/lib/AST/Type.cpp
index cbead120cbffc..2f6d8c1aa59ce 100644
--- a/clang/lib/AST/Type.cpp
+++ b/clang/lib/AST/Type.cpp
@@ -2787,7 +2787,7 @@ QualType Type::getRVVEltType(const ASTContext &Ctx) const {
bool QualType::isPODType(const ASTContext &Context) const {
if (Context.getLangOpts().HLSL &&
- getTypePtr()->isHLSLStandardRecordOrArrayOf())
+ getTypePtr()->isHLSLStandardLayoutRecordOrArrayOf())
return true;
// C++11 has a more relaxed definition of POD.
@@ -5533,11 +5533,11 @@ bool Type::isHLSLIntangibleType() const {
return RD->isHLSLIntangible();
}
-bool Type::isHLSLStandardRecordOrArrayOf() const {
+bool Type::isHLSLStandardLayoutRecordOrArrayOf() const {
const Type *BaseTy = getBaseElementTypeUnsafe();
if (const auto *RD =
dyn_cast_or_null<CXXRecordDecl>(BaseTy->getAsRecordDecl())) {
- if (!RD->isHLSLBuiltinRecord())
+ if (!RD->isHLSLBuiltinRecord() && RD->isStandardLayout())
return true;
}
return false;
diff --git a/clang/lib/CodeGen/CGExprAgg.cpp b/clang/lib/CodeGen/CGExprAgg.cpp
index b70aa2bf33b9b..8540603c28e9a 100644
--- a/clang/lib/CodeGen/CGExprAgg.cpp
+++ b/clang/lib/CodeGen/CGExprAgg.cpp
@@ -881,7 +881,11 @@ void AggExprEmitter::VisitCastExpr(CastExpr *E) {
// Create a temporary for the derived record, switch it out with the current
// Dest slot, and emit the derived value.
QualType DerivedTy = E->getSubExpr()->getType();
- AggValueSlot DerivedTmpSlot = CGF.CreateAggTemp(DerivedTy, "tmp");
+ RawAddress DerivedAddr = CGF.CreateMemTempWithoutCast(DerivedTy);
+ AggValueSlot DerivedTmpSlot = AggValueSlot::forAddr(
+ DerivedAddr, DerivedTy.getQualifiers(), AggValueSlot::IsNotDestructed,
+ AggValueSlot::DoesNotNeedGCBarriers, AggValueSlot::IsNotAliased,
+ AggValueSlot::DoesNotOverlap);
AggValueSlot DestBaseSlot = Dest;
Dest = DerivedTmpSlot;
>From 280781b7517fb25e34e27d57adad588df6b927c7 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 11 Jun 2026 21:36:15 -0700
Subject: [PATCH 20/22] refactor condition to keep original formatting
---
clang/include/clang/AST/DeclCXX.h | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index 868a9bb52215c..db05fa38818be 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -893,10 +893,11 @@ class CXXRecordDecl : public RecordDecl {
/// constructor or if any existing special member function inhibits this.
bool needsImplicitMoveConstructor() const {
return !(data().DeclaredSpecialMembers & SMF_MoveConstructor) &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord()) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveAssignment() && !hasUserDeclaredDestructor() &&
- (!getLangOpts().HLSL || isHLSLBuiltinRecord());
+ !hasUserDeclaredMoveAssignment() &&
+ !hasUserDeclaredDestructor();
}
/// Determine whether we need to eagerly declare a defaulted move
>From 99dc19dd77f673384e2452e9390ffaa3c7b5bc85 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Thu, 11 Jun 2026 22:12:51 -0700
Subject: [PATCH 21/22] two more condition refactoring to keep original
formatting and changes to a minimum + comment update
---
clang/include/clang/AST/DeclCXX.h | 27 ++++++++++++++-------------
clang/lib/Sema/SemaOverload.cpp | 2 +-
2 files changed, 15 insertions(+), 14 deletions(-)
diff --git a/clang/include/clang/AST/DeclCXX.h b/clang/include/clang/AST/DeclCXX.h
index db05fa38818be..28d171253dc03 100644
--- a/clang/include/clang/AST/DeclCXX.h
+++ b/clang/include/clang/AST/DeclCXX.h
@@ -763,16 +763,16 @@ class CXXRecordDecl : public RecordDecl {
/// this class.
///
/// This value is used for lazy creation of default constructors.
- bool needsImplicitDefaultConstructor() const {
- return ((!data().UserDeclaredConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
- // FIXME: Proposed fix to core wording issue: if a class inherits
- // a default constructor and doesn't explicitly declare one, one
- // is declared implicitly.
- (data().HasInheritedDefaultConstructor &&
- !(data().DeclaredSpecialMembers & SMF_DefaultConstructor))) &&
- (!getLangOpts().HLSL || isHLSLBuiltinRecord());
+ bool needsImplicitDefaultConstructor() const {
+ return (!getLangOpts().HLSL || isHLSLBuiltinRecord()) &&
+ ((!data().UserDeclaredConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor) &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable())) ||
+ // FIXME: Proposed fix to core wording issue: if a class inherits
+ // a default constructor and doesn't explicitly declare one, one
+ // is declared implicitly.
+ (data().HasInheritedDefaultConstructor &&
+ !(data().DeclaredSpecialMembers & SMF_DefaultConstructor)));
}
/// Determine whether this class has any user-declared constructors.
@@ -986,11 +986,12 @@ class CXXRecordDecl : public RecordDecl {
/// this.
bool needsImplicitMoveAssignment() const {
return !(data().DeclaredSpecialMembers & SMF_MoveAssignment) &&
+ (!getLangOpts().HLSL || isHLSLBuiltinRecord()) &&
!hasUserDeclaredCopyConstructor() &&
!hasUserDeclaredCopyAssignment() &&
- !hasUserDeclaredMoveConstructor() && !hasUserDeclaredDestructor() &&
- (!isLambda() || lambdaIsDefaultConstructibleAndAssignable()) &&
- (!getLangOpts().HLSL || isHLSLBuiltinRecord());
+ !hasUserDeclaredMoveConstructor() &&
+ !hasUserDeclaredDestructor() &&
+ (!isLambda() || lambdaIsDefaultConstructibleAndAssignable());
}
/// Determine whether we need to eagerly declare a move assignment
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 4ee935e7e6443..1960a917f52cb 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -15531,7 +15531,7 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
// problems. So we do it this way, which pretty much follows what GCC does.
// Note that we go the traditional code path for compound assignment forms.
// In HLSL, user-defined structs/classes do not have constructors or
- // overloadable operators, so we can take this shortcut too.
+ // overloadable assignment operators, so we can take this shortcut too.
const Type *LHSTy = Args[0]->getType().getTypePtr();
if (Opc == BO_Assign &&
(!LHSTy->isOverloadableType() ||
>From 0e5b54ac5a18b1e91b7799523cca164e0b922119 Mon Sep 17 00:00:00 2001
From: Helena Kotas <hekotas at microsoft.com>
Date: Fri, 12 Jun 2026 19:13:28 -0700
Subject: [PATCH 22/22] initialize IsHLSLBuiltinRecord field
---
clang/lib/AST/DeclCXX.cpp | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/clang/lib/AST/DeclCXX.cpp b/clang/lib/AST/DeclCXX.cpp
index fc8a15287f438..ce4ba971a4631 100644
--- a/clang/lib/AST/DeclCXX.cpp
+++ b/clang/lib/AST/DeclCXX.cpp
@@ -109,9 +109,10 @@ CXXRecordDecl::DefinitionData::DefinitionData(CXXRecordDecl *D)
ImplicitCopyAssignmentHasConstParam(true),
HasDeclaredCopyConstructorWithConstParam(false),
HasDeclaredCopyAssignmentWithConstParam(false),
- IsAnyDestructorNoReturn(false), IsHLSLIntangible(false), IsPFPType(false),
- IsLambda(false), IsParsingBaseSpecifiers(false),
- ComputedVisibleConversions(false), HasODRHash(false), Definition(D) {}
+ IsAnyDestructorNoReturn(false), IsHLSLIntangible(false),
+ IsHLSLBuiltinRecord(false), IsPFPType(false), IsLambda(false),
+ IsParsingBaseSpecifiers(false), ComputedVisibleConversions(false),
+ HasODRHash(false), Definition(D) {}
CXXBaseSpecifier *CXXRecordDecl::DefinitionData::getBasesSlowCase() const {
return Bases.get(Definition->getASTContext().getExternalSource());
More information about the cfe-commits
mailing list