[CIR] Upstream minimal support for structure types (#135105)

This change adds minimal support for structure types. To keep the
initial change small, only incomplete declarations are being supported
in this patch. More complete support will follow.

Author: andykaylor

clang/include/clang/CIR/Dialect/IR/CIRTypes.h

@@ -20,6 +20,10 @@
 
 namespace cir {
 
+namespace detail {
+struct RecordTypeStorage;
+} // namespace detail
+
 bool isAnyFloatingPointType(mlir::Type t);
 bool isFPOrFPVectorTy(mlir::Type);
 

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

@@ -400,13 +400,122 @@ def VoidPtr : Type<
       "cir::VoidType::get($_builder.getContext()))"> {
 }
 
+//===----------------------------------------------------------------------===//
+// RecordType
+//
+// The base type for all RecordDecls.
+//===----------------------------------------------------------------------===//
+
+def CIR_RecordType : CIR_Type<"Record", "record",
+    [
+      DeclareTypeInterfaceMethods<DataLayoutTypeInterface>,
+      MutableType,
+    ]> {
+  let summary = "CIR record type";
+  let description = [{
+    Each unique clang::RecordDecl is mapped to a `cir.record` and any object in
+    C/C++ that has a struct or class type will have a `cir.record` in CIR.
+
+    There are three possible formats for this type:
+
+     - Identified and complete records: unique name and a known body.
+     - Identified and incomplete records: unique name and unknown body.
+     - Anonymous records: no name and a known body.
+
+    Identified records are uniqued by their name, and anonymous records are
+    uniqued by their body. This means that two anonymous records with the same
+    body will be the same type, and two identified records with the same name
+    will be the same type. Attempting to build a record with an existing name,
+    but a different body will result in an error.
+
+    A few examples:
+
+    ```mlir
+        !complete = !cir.record<struct "complete" {!cir.int<u, 8>}>
+        !incomplete = !cir.record<struct "incomplete" incomplete>
+        !anonymous = !cir.record<struct {!cir.int<u, 8>}>
+    ```
+
+    Incomplete records are mutable, meaning they can be later completed with a
+    body automatically updating in place every type in the code that uses the
+    incomplete record. Mutability allows for recursive types to be represented,
+    meaning the record can have members that refer to itself. This is useful for
+    representing recursive records and is implemented through a special syntax.
+    In the example below, the `Node` record has a member that is a pointer to a
+    `Node` record:
+
+    ```mlir
+        !s = !cir.record<struct "Node" {!cir.ptr<!cir.record<struct "Node">>}>
+    ```
+  }];
+
+  let parameters = (ins
+    OptionalArrayRefParameter<"mlir::Type">:$members,
+    OptionalParameter<"mlir::StringAttr">:$name,
+    "bool":$incomplete,
+    "bool":$packed,
+    "bool":$padded,
+    "RecordType::RecordKind":$kind
+  );
+
+  // StorageClass is defined in C++ for mutability.
+  let storageClass = "RecordTypeStorage";
+  let genStorageClass = 0;
+
+  let skipDefaultBuilders = 1;
+  let genVerifyDecl = 1;
+
+  let builders = [
+    // Create an identified and incomplete record type.
+    TypeBuilder<(ins
+      "mlir::StringAttr":$name,
+      "RecordKind":$kind
+    ), [{
+      return $_get($_ctxt, /*members=*/llvm::ArrayRef<Type>{}, name,
+                         /*incomplete=*/true, /*packed=*/false,
+                         /*padded=*/false, kind);
+    }]>];
+
+  let extraClassDeclaration = [{
+    using Base::verifyInvariants;
+
+    enum RecordKind : uint32_t { Struct, Union };
+
+    bool isStruct() const { return getKind() == RecordKind::Struct; };
+    bool isUnion() const { return getKind() == RecordKind::Union; };
+    bool isComplete() const { return !isIncomplete(); };
+    bool isIncomplete() const;
+
+    size_t getNumElements() const { return getMembers().size(); };
+    std::string getKindAsStr() {
+      switch (getKind()) {
+      case RecordKind::Union:
+        return "union";
+      case RecordKind::Struct:
+        return "struct";
+      }
+      llvm_unreachable("Invalid value for RecordType::getKind()");
+    }
+    std::string getPrefixedName() {
+      return getKindAsStr() + "." + getName().getValue().str();
+    }
+  }];
+
+  let hasCustomAssemblyFormat = 1;
+}
+
+// Note CIRRecordType is used instead of CIR_RecordType
+// because of tablegen conflicts.
+def CIRRecordType : Type<
+  CPred<"::mlir::isa<::cir::RecordType>($_self)">, "CIR record type">;
+
 //===----------------------------------------------------------------------===//
 // Global type constraints
 //===----------------------------------------------------------------------===//
 
 def CIR_AnyType : AnyTypeOf<[
   CIR_VoidType, CIR_BoolType, CIR_ArrayType, CIR_IntType, CIR_AnyFloat,
-  CIR_PointerType, CIR_FuncType
+  CIR_PointerType, CIR_FuncType, CIR_RecordType
 ]>;
 
 #endif // MLIR_CIR_DIALECT_CIR_TYPES

clang/include/clang/CIR/Dialect/IR/CIRTypesDetails.h

@@ -0,0 +1,118 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains implementation details, such as storage structures, of
+// CIR dialect types.
+//
+//===----------------------------------------------------------------------===//
+#ifndef CIR_DIALECT_IR_CIRTYPESDETAILS_H
+#define CIR_DIALECT_IR_CIRTYPESDETAILS_H
+
+#include "mlir/IR/BuiltinAttributes.h"
+#include "mlir/Support/LogicalResult.h"
+#include "clang/CIR/Dialect/IR/CIRTypes.h"
+#include "llvm/ADT/Hashing.h"
+
+namespace cir {
+namespace detail {
+
+//===----------------------------------------------------------------------===//
+// CIR RecordTypeStorage
+//===----------------------------------------------------------------------===//
+
+/// Type storage for CIR record types.
+struct RecordTypeStorage : public mlir::TypeStorage {
+  struct KeyTy {
+    llvm::ArrayRef<mlir::Type> members;
+    mlir::StringAttr name;
+    bool incomplete;
+    bool packed;
+    bool padded;
+    RecordType::RecordKind kind;
+
+    KeyTy(llvm::ArrayRef<mlir::Type> members, mlir::StringAttr name,
+          bool incomplete, bool packed, bool padded,
+          RecordType::RecordKind kind)
+        : members(members), name(name), incomplete(incomplete), packed(packed),
+          padded(padded), kind(kind) {}
+  };
+
+  llvm::ArrayRef<mlir::Type> members;
+  mlir::StringAttr name;
+  bool incomplete;
+  bool packed;
+  bool padded;
+  RecordType::RecordKind kind;
+
+  RecordTypeStorage(llvm::ArrayRef<mlir::Type> members, mlir::StringAttr name,
+                    bool incomplete, bool packed, bool padded,
+                    RecordType::RecordKind kind)
+      : members(members), name(name), incomplete(incomplete), packed(packed),
+        padded(padded), kind(kind) {
+    assert(name || !incomplete && "Incomplete records must have a name");
+  }
+
+  KeyTy getAsKey() const {
+    return KeyTy(members, name, incomplete, packed, padded, kind);
+  }
+
+  bool operator==(const KeyTy &key) const {
+    if (name)
+      return (name == key.name) && (kind == key.kind);
+    return std::tie(members, name, incomplete, packed, padded, kind) ==
+           std::tie(key.members, key.name, key.incomplete, key.packed,
+                    key.padded, key.kind);
+  }
+
+  static llvm::hash_code hashKey(const KeyTy &key) {
+    if (key.name)
+      return llvm::hash_combine(key.name, key.kind);
+    return llvm::hash_combine(key.members, key.incomplete, key.packed,
+                              key.padded, key.kind);
+  }
+
+  static RecordTypeStorage *construct(mlir::TypeStorageAllocator &allocator,
+                                      const KeyTy &key) {
+    return new (allocator.allocate<RecordTypeStorage>())
+        RecordTypeStorage(allocator.copyInto(key.members), key.name,
+                          key.incomplete, key.packed, key.padded, key.kind);
+  }
+
+  /// Mutates the members and attributes an identified record.
+  ///
+  /// Once a record is mutated, it is marked as complete, preventing further
+  /// mutations. Anonymous records are always complete and cannot be mutated.
+  /// This method does not fail if a mutation of a complete record does not
+  /// change the record.
+  llvm::LogicalResult mutate(mlir::TypeStorageAllocator &allocator,
+                             llvm::ArrayRef<mlir::Type> members, bool packed,
+                             bool padded) {
+    // Anonymous records cannot mutate.
+    if (!name)
+      return llvm::failure();
+
+    // Mutation of complete records are allowed if they change nothing.
+    if (!incomplete)
+      return mlir::success((this->members == members) &&
+                           (this->packed == packed) &&
+                           (this->padded == padded));
+
+    // Mutate incomplete record.
+    this->members = allocator.copyInto(members);
+    this->packed = packed;
+    this->padded = padded;
+
+    incomplete = false;
+    return llvm::success();
+  }
+};
+
+} // namespace detail
+} // namespace cir
+
+#endif // CIR_DIALECT_IR_CIRTYPESDETAILS_H

clang/include/clang/CIR/MissingFeatures.h

@@ -101,6 +101,9 @@ struct MissingFeatures {
   static bool mayHaveIntegerOverflow() { return false; }
   static bool shouldReverseUnaryCondOnBoolExpr() { return false; }
 
+  // RecordType
+  static bool recordTypeLayoutInfo() { return false; }
+
   // Misc
   static bool cxxABI() { return false; }
   static bool tryEmitAsConstant() { return false; }

clang/lib/CIR/CodeGen/CIRGenBuilder.h

@@ -20,11 +20,24 @@ namespace clang::CIRGen {
 
 class CIRGenBuilderTy : public cir::CIRBaseBuilderTy {
   const CIRGenTypeCache &typeCache;
+  llvm::StringMap<unsigned> recordNames;
 
 public:
   CIRGenBuilderTy(mlir::MLIRContext &mlirContext, const CIRGenTypeCache &tc)
       : CIRBaseBuilderTy(mlirContext), typeCache(tc) {}
 
+  std::string getUniqueAnonRecordName() { return getUniqueRecordName("anon"); }
+
+  std::string getUniqueRecordName(const std::string &baseName) {
+    auto it = recordNames.find(baseName);
+    if (it == recordNames.end()) {
+      recordNames[baseName] = 0;
+      return baseName;
+    }
+
+    return baseName + "." + std::to_string(recordNames[baseName]++);
+  }
+
   cir::LongDoubleType getLongDoubleTy(const llvm::fltSemantics &format) const {
     if (&format == &llvm::APFloat::IEEEdouble())
       return cir::LongDoubleType::get(getContext(), typeCache.DoubleTy);
@@ -37,6 +50,33 @@ class CIRGenBuilderTy : public cir::CIRBaseBuilderTy {
     llvm_unreachable("Unsupported format for long double");
   }
 
+  /// Get a CIR record kind from a AST declaration tag.
+  cir::RecordType::RecordKind getRecordKind(const clang::TagTypeKind kind) {
+    switch (kind) {
+    case clang::TagTypeKind::Class:
+    case clang::TagTypeKind::Struct:
+      return cir::RecordType::Struct;
+    case clang::TagTypeKind::Union:
+      return cir::RecordType::Union;
+    case clang::TagTypeKind::Interface:
+      llvm_unreachable("interface records are NYI");
+    case clang::TagTypeKind::Enum:
+      llvm_unreachable("enums are not records");
+    }
+  }
+
+  /// Get an incomplete CIR struct type. If we have a complete record
+  /// declaration, we may create an incomplete type and then add the
+  /// members, so \p rd here may be complete.
+  cir::RecordType getIncompleteRecordTy(llvm::StringRef name,
+                                        const clang::RecordDecl *rd) {
+    const mlir::StringAttr nameAttr = getStringAttr(name);
+    cir::RecordType::RecordKind kind = cir::RecordType::RecordKind::Struct;
+    if (rd)
+      kind = getRecordKind(rd->getTagKind());
+    return getType<cir::RecordType>(nameAttr, kind);
+  }
+
   bool isSized(mlir::Type ty) {
     if (mlir::isa<cir::PointerType, cir::ArrayType, cir::BoolType,
                   cir::IntType>(ty))

clang/lib/CIR/CodeGen/CIRGenDecl.cpp

@@ -260,6 +260,9 @@ void CIRGenFunction::emitExprAsInit(const Expr *init, const ValueDecl *d,
 
 void CIRGenFunction::emitDecl(const Decl &d) {
   switch (d.getKind()) {
+  case Decl::Record: // struct/union/class X;
+    assert(!cir::MissingFeatures::generateDebugInfo());
+    return;
   case Decl::Var: {
     const VarDecl &vd = cast<VarDecl>(d);
     assert(vd.isLocalVarDecl() &&
@@ -274,7 +277,9 @@ void CIRGenFunction::emitDecl(const Decl &d) {
     emitOpenACCRoutine(cast<OpenACCRoutineDecl>(d));
     return;
   default:
-    cgm.errorNYI(d.getSourceRange(), "emitDecl: unhandled decl type");
+    cgm.errorNYI(d.getSourceRange(),
+                 std::string("emitDecl: unhandled decl type: ") +
+                     d.getDeclKindName());
   }
 }
 

clang/lib/CIR/CodeGen/CIRGenModule.cpp

@@ -589,6 +589,11 @@ void CIRGenModule::emitTopLevelDecl(Decl *decl) {
   case Decl::OpenACCDeclare:
     emitGlobalOpenACCDecl(cast<OpenACCDeclareDecl>(decl));
     break;
+
+  case Decl::Record:
+  case Decl::CXXRecord:
+    assert(!cir::MissingFeatures::generateDebugInfo());
+    break;
   }
 }