merge main into amd-staging (llvm#4135)

Author: z1-cciauto

clang/docs/ReleaseNotes.rst

@@ -519,6 +519,8 @@ Improvements to Clang's diagnostics
   Moved the warning for a missing (though implied) attribute on a redeclaration into this group.
   Added a new warning in this group for the case where the attribute is missing/implicit on
   an override of a virtual method.
+- Implemented diagnostics when retrieving the tuple size for types where its specialization of `std::tuple_size`
+  produces an invalid size (either negative or greater than the implementation limit). (#GH159563)
 - Fixed fix-it hint for fold expressions. Clang now correctly places the suggested right
   parenthesis when diagnosing malformed fold expressions. (#GH151787)
 - Added fix-it hint for when scoped enumerations require explicit conversions for binary operations. (#GH24265)
@@ -604,8 +606,8 @@ Bug Fixes in This Version
   and vector of 4 ``float`` values. (#GH155405)
 - Fixed inconsistent shadow warnings for lambda capture of structured bindings.
   Previously, ``[val = val]`` (regular parameter) produced no warnings with ``-Wshadow``
-  while ``[a = a]`` (where ``a`` is from ``auto [a, b] = std::make_pair(1, 2)``) 
-  incorrectly produced warnings. Both cases now consistently show no warnings with 
+  while ``[a = a]`` (where ``a`` is from ``auto [a, b] = std::make_pair(1, 2)``)
+  incorrectly produced warnings. Both cases now consistently show no warnings with
   ``-Wshadow`` and show uncaptured-local warnings with ``-Wshadow-all``. (#GH68605)
 - Fixed a failed assertion with a negative limit parameter value inside of
   ``__has_embed``. (#GH157842)

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -638,6 +638,9 @@ def err_decomp_decl_std_tuple_element_not_specialized : Error<
 def err_decomp_decl_std_tuple_size_not_constant : Error<
   "cannot decompose this type; 'std::tuple_size<%0>::value' "
   "is not a valid integral constant expression">;
+def err_decomp_decl_std_tuple_size_invalid
+    : Error<"cannot decompose this type; 'std::tuple_size<%0>::value' "
+            "is not a valid size: %1">;
 def note_in_binding_decl_init : Note<
   "in implicit initialization of binding declaration %0">;
 def err_arg_is_not_destructurable : Error<

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -4131,6 +4131,45 @@ def CIR_ThrowOp : CIR_Op<"throw"> {
 // Atomic operations
 //===----------------------------------------------------------------------===//
 
+def CIR_AtomicXchg : CIR_Op<"atomic.xchg", [
+  AllTypesMatch<["result", "val"]>,
+  TypesMatchWith<"type of 'val' must match the pointee type of 'ptr'",
+    "ptr", "val", "mlir::cast<cir::PointerType>($_self).getPointee()">
+]> {
+  let summary = "Atomic exchange";
+  let description = [{
+    C/C++ atomic exchange operation. This operation implements the C/C++
+    builtin function `__atomic_exchange`, `__atomic_exchange_n`, and
+    `__c11_atomic_exchange`.
+
+    This operation takes two arguments: a pointer `ptr` and a value `val`. The
+    operation atomically replaces the value of the object pointed-to by `ptr`
+    with `val`, and returns the original value of the object.
+
+    Example:
+
+    ```mlir
+    %res = cir.atomic.xchg(%ptr : !cir.ptr<!u64i>,
+                           %val : !u64i,
+                           seq_cst) : !u64i
+    ```
+  }];
+
+  let results = (outs CIR_AnyType:$result);
+  let arguments = (ins
+    Arg<CIR_PointerType, "", [MemRead, MemWrite]>:$ptr,
+    CIR_AnyType:$val,
+    Arg<CIR_MemOrder, "memory order">:$mem_order,
+    UnitAttr:$is_volatile
+  );
+
+  let assemblyFormat = [{
+    $mem_order (`volatile` $is_volatile^)?
+    $ptr `,` $val
+    `:` qualified(type($ptr)) `->` type($result) attr-dict
+  }];
+}
+
 def CIR_AtomicCmpXchg : CIR_Op<"atomic.cmpxchg", [
   AllTypesMatch<["old", "expected", "desired"]>
 ]> {

clang/lib/AST/ASTContext.cpp

@@ -1712,6 +1712,9 @@ ASTContext::findPointerAuthContent(QualType T) const {
   assert(isPointerAuthenticationAvailable());
 
   T = T.getCanonicalType();
+  if (T->isDependentType())
+    return PointerAuthContent::None;
+
   if (T.hasAddressDiscriminatedPointerAuth())
     return PointerAuthContent::AddressDiscriminatedData;
   const RecordDecl *RD = T->getAsRecordDecl();

clang/lib/CIR/CodeGen/CIRGenAtomic.cpp

@@ -341,6 +341,7 @@ static void emitAtomicOp(CIRGenFunction &cgf, AtomicExpr *expr, Address dest,
   }
 
   assert(!cir::MissingFeatures::atomicSyncScopeID());
+  llvm::StringRef opName;
 
   CIRGenBuilderTy &builder = cgf.getBuilder();
   mlir::Location loc = cgf.getLoc(expr->getSourceRange());
@@ -400,6 +401,12 @@ static void emitAtomicOp(CIRGenFunction &cgf, AtomicExpr *expr, Address dest,
     return;
   }
 
+  case AtomicExpr::AO__c11_atomic_exchange:
+  case AtomicExpr::AO__atomic_exchange_n:
+  case AtomicExpr::AO__atomic_exchange:
+    opName = cir::AtomicXchg::getOperationName();
+    break;
+
   case AtomicExpr::AO__opencl_atomic_init:
 
   case AtomicExpr::AO__hip_atomic_compare_exchange_strong:
@@ -421,11 +428,8 @@ static void emitAtomicOp(CIRGenFunction &cgf, AtomicExpr *expr, Address dest,
   case AtomicExpr::AO__scoped_atomic_store:
   case AtomicExpr::AO__scoped_atomic_store_n:
 
-  case AtomicExpr::AO__c11_atomic_exchange:
   case AtomicExpr::AO__hip_atomic_exchange:
   case AtomicExpr::AO__opencl_atomic_exchange:
-  case AtomicExpr::AO__atomic_exchange_n:
-  case AtomicExpr::AO__atomic_exchange:
   case AtomicExpr::AO__scoped_atomic_exchange_n:
   case AtomicExpr::AO__scoped_atomic_exchange:
 
@@ -503,8 +507,23 @@ static void emitAtomicOp(CIRGenFunction &cgf, AtomicExpr *expr, Address dest,
 
   case AtomicExpr::AO__atomic_clear:
     cgf.cgm.errorNYI(expr->getSourceRange(), "emitAtomicOp: expr op NYI");
-    break;
+    return;
   }
+
+  assert(!opName.empty() && "expected operation name to build");
+  mlir::Value loadVal1 = builder.createLoad(loc, val1);
+
+  SmallVector<mlir::Value> atomicOperands = {ptr.getPointer(), loadVal1};
+  SmallVector<mlir::Type> atomicResTys = {loadVal1.getType()};
+  mlir::Operation *rmwOp = builder.create(loc, builder.getStringAttr(opName),
+                                          atomicOperands, atomicResTys);
+
+  rmwOp->setAttr("mem_order", orderAttr);
+  if (expr->isVolatile())
+    rmwOp->setAttr("is_volatile", builder.getUnitAttr());
+
+  mlir::Value result = rmwOp->getResult(0);
+  builder.createStore(loc, result, dest);
 }
 
 static bool isMemOrderValid(uint64_t order, bool isStore, bool isLoad) {
@@ -572,6 +591,11 @@ RValue CIRGenFunction::emitAtomicExpr(AtomicExpr *e) {
     val1 = emitPointerWithAlignment(e->getVal1());
     break;
 
+  case AtomicExpr::AO__atomic_exchange:
+    val1 = emitPointerWithAlignment(e->getVal1());
+    dest = emitPointerWithAlignment(e->getVal2());
+    break;
+
   case AtomicExpr::AO__atomic_compare_exchange:
   case AtomicExpr::AO__atomic_compare_exchange_n:
   case AtomicExpr::AO__c11_atomic_compare_exchange_weak:
@@ -590,7 +614,9 @@ RValue CIRGenFunction::emitAtomicExpr(AtomicExpr *e) {
       isWeakExpr = e->getWeak();
     break;
 
+  case AtomicExpr::AO__atomic_exchange_n:
   case AtomicExpr::AO__atomic_store_n:
+  case AtomicExpr::AO__c11_atomic_exchange:
   case AtomicExpr::AO__c11_atomic_store:
     val1 = emitValToTemp(*this, e->getVal1());
     break;

clang/lib/CIR/CodeGen/CIRGenClass.cpp

@@ -639,12 +639,22 @@ void CIRGenFunction::emitCXXAggrConstructorCall(
   // are probably legitimate places where we could assume that this
   // doesn't happen, but it's not clear that it's worth it.
 
+  auto arrayTy = mlir::cast<cir::ArrayType>(arrayBase.getElementType());
+  mlir::Type elementType = arrayTy.getElementType();
+
+  // This might be a multi-dimensional array. Find the innermost element type.
+  while (auto maybeArrayTy = mlir::dyn_cast<cir::ArrayType>(elementType))
+    elementType = maybeArrayTy.getElementType();
+  cir::PointerType ptrToElmType = builder.getPointerTo(elementType);
+
   // Optimize for a constant count.
   if (auto constantCount = numElements.getDefiningOp<cir::ConstantOp>()) {
     if (auto constIntAttr = constantCount.getValueAttr<cir::IntAttr>()) {
       // Just skip out if the constant count is zero.
       if (constIntAttr.getUInt() == 0)
         return;
+
+      arrayTy = cir::ArrayType::get(elementType, constIntAttr.getUInt());
       // Otherwise, emit the check.
     }
 
@@ -655,10 +665,6 @@ void CIRGenFunction::emitCXXAggrConstructorCall(
     cgm.errorNYI(e->getSourceRange(), "dynamic-length array expression");
   }
 
-  auto arrayTy = mlir::cast<cir::ArrayType>(arrayBase.getElementType());
-  mlir::Type elementType = arrayTy.getElementType();
-  cir::PointerType ptrToElmType = builder.getPointerTo(elementType);
-
   // Tradional LLVM codegen emits a loop here. CIR lowers to a loop as part of
   // LoweringPrepare.
 

clang/lib/CIR/CodeGen/CIRGenDecl.cpp

@@ -618,6 +618,7 @@ void CIRGenFunction::emitDecl(const Decl &d, bool evaluateConditionDecl) {
   case Decl::MSGuid: // __declspec(uuid("..."))
   case Decl::TemplateParamObject:
   case Decl::OMPThreadPrivate:
+  case Decl::OMPGroupPrivate:
   case Decl::OMPAllocate:
   case Decl::OMPCapturedExpr:
   case Decl::OMPRequires:
@@ -667,7 +668,6 @@ void CIRGenFunction::emitDecl(const Decl &d, bool evaluateConditionDecl) {
   case Decl::UsingPack:
   case Decl::OMPDeclareMapper:
   case Decl::OMPDeclareReduction:
-  case Decl::OMPGroupPrivate:
     cgm.errorNYI(d.getSourceRange(),
                  std::string("emitDecl: unhandled decl type: ") +
                      d.getDeclKindName());
@@ -713,10 +713,11 @@ void CIRGenFunction::pushDestroy(CleanupKind cleanupKind, Address addr,
 /// The array cannot be zero-length.
 ///
 /// \param begin - a type* denoting the first element of the array
-/// \param end - a type* denoting one past the end of the array
+/// \param numElements - the number of elements in the array
 /// \param elementType - the element type of the array
 /// \param destroyer - the function to call to destroy elements
-void CIRGenFunction::emitArrayDestroy(mlir::Value begin, mlir::Value end,
+void CIRGenFunction::emitArrayDestroy(mlir::Value begin,
+                                      mlir::Value numElements,
                                       QualType elementType,
                                       CharUnits elementAlign,
                                       Destroyer *destroyer) {
@@ -727,9 +728,24 @@ void CIRGenFunction::emitArrayDestroy(mlir::Value begin, mlir::Value end,
   mlir::Type cirElementType = convertTypeForMem(elementType);
   cir::PointerType ptrToElmType = builder.getPointerTo(cirElementType);
 
+  uint64_t size = 0;
+
+  // Optimize for a constant array size.
+  if (auto constantCount = numElements.getDefiningOp<cir::ConstantOp>()) {
+    if (auto constIntAttr = constantCount.getValueAttr<cir::IntAttr>())
+      size = constIntAttr.getUInt();
+  } else {
+    cgm.errorNYI(begin.getDefiningOp()->getLoc(),
+                 "dynamic-length array expression");
+  }
+
+  auto arrayTy = cir::ArrayType::get(cirElementType, size);
+  mlir::Value arrayOp = builder.createPtrBitcast(begin, arrayTy);
+
   // Emit the dtor call that will execute for every array element.
   cir::ArrayDtor::create(
-      builder, *currSrcLoc, begin, [&](mlir::OpBuilder &b, mlir::Location loc) {
+      builder, *currSrcLoc, arrayOp,
+      [&](mlir::OpBuilder &b, mlir::Location loc) {
         auto arg = b.getInsertionBlock()->addArgument(ptrToElmType, loc);
         Address curAddr = Address(arg, cirElementType, elementAlign);
         assert(!cir::MissingFeatures::dtorCleanups());
@@ -772,8 +788,7 @@ void CIRGenFunction::emitDestroy(Address addr, QualType type,
     return;
 
   mlir::Value begin = addr.getPointer();
-  mlir::Value end; // This will be used for future non-constant counts.
-  emitArrayDestroy(begin, end, type, elementAlign, destroyer);
+  emitArrayDestroy(begin, length, type, elementAlign, destroyer);
 
   // If the array destroy didn't use the length op, we can erase it.
   if (constantCount.use_empty())

clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp

@@ -878,10 +878,9 @@ class ScalarExprEmitter : public StmtVisitor<ScalarExprEmitter, mlir::Value> {
     }
 
     if (ty.UseExcessPrecision(cgf.getContext())) {
-      if (ty->getAs<VectorType>()) {
-        assert(!cir::MissingFeatures::vectorType());
-        cgf.cgm.errorNYI("getPromotionType: promotion to vector type");
-        return QualType();
+      if (auto *vt = ty->getAs<VectorType>()) {
+        unsigned numElements = vt->getNumElements();
+        return ctx.getVectorType(ctx.FloatTy, numElements, vt->getVectorKind());
       }
       return cgf.getContext().FloatTy;
     }
@@ -1063,8 +1062,20 @@ class ScalarExprEmitter : public StmtVisitor<ScalarExprEmitter, mlir::Value> {
 
   mlir::Value VisitBinLAnd(const clang::BinaryOperator *e) {
     if (e->getType()->isVectorType()) {
-      assert(!cir::MissingFeatures::vectorType());
-      return {};
+      mlir::Location loc = cgf.getLoc(e->getExprLoc());
+      auto vecTy = mlir::cast<cir::VectorType>(cgf.convertType(e->getType()));
+      mlir::Value zeroValue = builder.getNullValue(vecTy.getElementType(), loc);
+      SmallVector<mlir::Value, 16> elements(vecTy.getSize(), zeroValue);
+      auto zeroVec = cir::VecCreateOp::create(builder, loc, vecTy, elements);
+
+      mlir::Value lhs = Visit(e->getLHS());
+      mlir::Value rhs = Visit(e->getRHS());
+
+      auto cmpOpKind = cir::CmpOpKind::ne;
+      lhs = cir::VecCmpOp::create(builder, loc, vecTy, cmpOpKind, lhs, zeroVec);
+      rhs = cir::VecCmpOp::create(builder, loc, vecTy, cmpOpKind, rhs, zeroVec);
+      mlir::Value vecOr = builder.createAnd(loc, lhs, rhs);
+      return builder.createIntCast(vecOr, vecTy);
     }
 
     assert(!cir::MissingFeatures::instrumentation());
@@ -2344,4 +2355,4 @@ mlir::Value CIRGenFunction::emitScalarPrePostIncDec(const UnaryOperator *e,
                                                     bool isPre) {
   return ScalarExprEmitter(*this, builder)
       .emitScalarPrePostIncDec(e, lv, kind, isPre);
-}
+}

clang/lib/CIR/CodeGen/CIRGenFunction.h

@@ -1064,7 +1064,7 @@ class CIRGenFunction : public CIRGenTypeCache {
   /// even if no aggregate location is provided.
   RValue emitAnyExprToTemp(const clang::Expr *e);
 
-  void emitArrayDestroy(mlir::Value begin, mlir::Value end,
+  void emitArrayDestroy(mlir::Value begin, mlir::Value numElements,
                         QualType elementType, CharUnits elementAlign,
                         Destroyer *destroyer);
 

clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp

@@ -717,6 +717,17 @@ mlir::LogicalResult CIRToLLVMAtomicCmpXchgLowering::matchAndRewrite(
   return mlir::success();
 }
 
+mlir::LogicalResult CIRToLLVMAtomicXchgLowering::matchAndRewrite(
+    cir::AtomicXchg op, OpAdaptor adaptor,
+    mlir::ConversionPatternRewriter &rewriter) const {
+  assert(!cir::MissingFeatures::atomicSyncScopeID());
+  mlir::LLVM::AtomicOrdering llvmOrder = getLLVMMemOrder(adaptor.getMemOrder());
+  rewriter.replaceOpWithNewOp<mlir::LLVM::AtomicRMWOp>(
+      op, mlir::LLVM::AtomicBinOp::xchg, adaptor.getPtr(), adaptor.getVal(),
+      llvmOrder);
+  return mlir::success();
+}
+
 mlir::LogicalResult CIRToLLVMBitClrsbOpLowering::matchAndRewrite(
     cir::BitClrsbOp op, OpAdaptor adaptor,
     mlir::ConversionPatternRewriter &rewriter) const {