[CIR] Upstream ArraySubscriptExpr for fixed size array (#134536)

This change adds ArraySubscriptExpr for fixed size ArrayType

Issue #130197

Author: AmrDeveloper

clang/include/clang/CIR/MissingFeatures.h

@@ -141,6 +141,8 @@ struct MissingFeatures {
   static bool mangledNames() { return false; }
   static bool setDLLStorageClass() { return false; }
   static bool openMP() { return false; }
+  static bool emitCheckedInBoundsGEP() { return false; }
+  static bool preservedAccessIndexRegion() { return false; }
 
   // Missing types
   static bool dataMemberType() { return false; }

clang/lib/CIR/CodeGen/CIRGenBuilder.cpp

@@ -0,0 +1,40 @@
+//===----------------------------------------------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+#include "CIRGenBuilder.h"
+
+using namespace clang::CIRGen;
+
+mlir::Value CIRGenBuilderTy::maybeBuildArrayDecay(mlir::Location loc,
+                                                  mlir::Value arrayPtr,
+                                                  mlir::Type eltTy) {
+  const auto arrayPtrTy = mlir::cast<cir::PointerType>(arrayPtr.getType());
+  const auto arrayTy = mlir::dyn_cast<cir::ArrayType>(arrayPtrTy.getPointee());
+
+  if (arrayTy) {
+    const cir::PointerType flatPtrTy = getPointerTo(arrayTy.getEltType());
+    return create<cir::CastOp>(loc, flatPtrTy, cir::CastKind::array_to_ptrdecay,
+                               arrayPtr);
+  }
+
+  assert(arrayPtrTy.getPointee() == eltTy &&
+         "flat pointee type must match original array element type");
+  return arrayPtr;
+}
+
+mlir::Value CIRGenBuilderTy::getArrayElement(mlir::Location arrayLocBegin,
+                                             mlir::Location arrayLocEnd,
+                                             mlir::Value arrayPtr,
+                                             mlir::Type eltTy, mlir::Value idx,
+                                             bool shouldDecay) {
+  mlir::Value basePtr = arrayPtr;
+  if (shouldDecay)
+    basePtr = maybeBuildArrayDecay(arrayLocBegin, arrayPtr, eltTy);
+  const mlir::Type flatPtrTy = basePtr.getType();
+  return create<cir::PtrStrideOp>(arrayLocEnd, flatPtrTy, basePtr, idx);
+}

clang/lib/CIR/CodeGen/CIRGenBuilder.h

@@ -198,6 +198,19 @@ class CIRGenBuilderTy : public cir::CIRBaseBuilderTy {
 
     return create<cir::BinOp>(loc, cir::BinOpKind::Div, lhs, rhs);
   }
+
+  /// Create a cir.ptr_stride operation to get access to an array element.
+  /// \p idx is the index of the element to access, \p shouldDecay is true if
+  /// the result should decay to a pointer to the element type.
+  mlir::Value getArrayElement(mlir::Location arrayLocBegin,
+                              mlir::Location arrayLocEnd, mlir::Value arrayPtr,
+                              mlir::Type eltTy, mlir::Value idx,
+                              bool shouldDecay);
+
+  /// Returns a decayed pointer to the first element of the array
+  /// pointed to by \p arrayPtr.
+  mlir::Value maybeBuildArrayDecay(mlir::Location loc, mlir::Value arrayPtr,
+                                   mlir::Type eltTy);
 };
 
 } // namespace clang::CIRGen

clang/lib/CIR/CodeGen/CIRGenExpr.cpp

@@ -12,6 +12,7 @@
 
 #include "Address.h"
 #include "CIRGenFunction.h"
+#include "CIRGenModule.h"
 #include "CIRGenValue.h"
 #include "mlir/IR/BuiltinAttributes.h"
 #include "clang/AST/Attr.h"
@@ -430,6 +431,143 @@ LValue CIRGenFunction::emitUnaryOpLValue(const UnaryOperator *e) {
   llvm_unreachable("Unknown unary operator kind!");
 }
 
+/// If the specified expr is a simple decay from an array to pointer,
+/// return the array subexpression.
+/// FIXME: this could be abstracted into a common AST helper.
+static const Expr *getSimpleArrayDecayOperand(const Expr *e) {
+  // If this isn't just an array->pointer decay, bail out.
+  const auto *castExpr = dyn_cast<CastExpr>(e);
+  if (!castExpr || castExpr->getCastKind() != CK_ArrayToPointerDecay)
+    return nullptr;
+
+  // If this is a decay from variable width array, bail out.
+  const Expr *subExpr = castExpr->getSubExpr();
+  if (subExpr->getType()->isVariableArrayType())
+    return nullptr;
+
+  return subExpr;
+}
+
+static cir::IntAttr getConstantIndexOrNull(mlir::Value idx) {
+  // TODO(cir): should we consider using MLIRs IndexType instead of IntegerAttr?
+  if (auto constantOp = dyn_cast<cir::ConstantOp>(idx.getDefiningOp()))
+    return mlir::dyn_cast<cir::IntAttr>(constantOp.getValue());
+  return {};
+}
+
+static CharUnits getArrayElementAlign(CharUnits arrayAlign, mlir::Value idx,
+                                      CharUnits eltSize) {
+  // If we have a constant index, we can use the exact offset of the
+  // element we're accessing.
+  const cir::IntAttr constantIdx = getConstantIndexOrNull(idx);
+  if (constantIdx) {
+    const CharUnits offset = constantIdx.getValue().getZExtValue() * eltSize;
+    return arrayAlign.alignmentAtOffset(offset);
+  }
+  // Otherwise, use the worst-case alignment for any element.
+  return arrayAlign.alignmentOfArrayElement(eltSize);
+}
+
+static QualType getFixedSizeElementType(const ASTContext &astContext,
+                                        const VariableArrayType *vla) {
+  QualType eltType;
+  do {
+    eltType = vla->getElementType();
+  } while ((vla = astContext.getAsVariableArrayType(eltType)));
+  return eltType;
+}
+
+static mlir::Value emitArraySubscriptPtr(CIRGenFunction &cgf,
+                                         mlir::Location beginLoc,
+                                         mlir::Location endLoc, mlir::Value ptr,
+                                         mlir::Type eltTy, mlir::Value idx,
+                                         bool shouldDecay) {
+  CIRGenModule &cgm = cgf.getCIRGenModule();
+  // TODO(cir): LLVM codegen emits in bound gep check here, is there anything
+  // that would enhance tracking this later in CIR?
+  assert(!cir::MissingFeatures::emitCheckedInBoundsGEP());
+  return cgm.getBuilder().getArrayElement(beginLoc, endLoc, ptr, eltTy, idx,
+                                          shouldDecay);
+}
+
+static Address emitArraySubscriptPtr(CIRGenFunction &cgf,
+                                     mlir::Location beginLoc,
+                                     mlir::Location endLoc, Address addr,
+                                     QualType eltType, mlir::Value idx,
+                                     mlir::Location loc, bool shouldDecay) {
+
+  // Determine the element size of the statically-sized base.  This is
+  // the thing that the indices are expressed in terms of.
+  if (const VariableArrayType *vla =
+          cgf.getContext().getAsVariableArrayType(eltType)) {
+    eltType = getFixedSizeElementType(cgf.getContext(), vla);
+  }
+
+  // We can use that to compute the best alignment of the element.
+  const CharUnits eltSize = cgf.getContext().getTypeSizeInChars(eltType);
+  const CharUnits eltAlign =
+      getArrayElementAlign(addr.getAlignment(), idx, eltSize);
+
+  assert(!cir::MissingFeatures::preservedAccessIndexRegion());
+  const mlir::Value eltPtr =
+      emitArraySubscriptPtr(cgf, beginLoc, endLoc, addr.getPointer(),
+                            addr.getElementType(), idx, shouldDecay);
+  const mlir::Type elementType = cgf.convertTypeForMem(eltType);
+  return Address(eltPtr, elementType, eltAlign);
+}
+
+LValue
+CIRGenFunction::emitArraySubscriptExpr(const clang::ArraySubscriptExpr *e) {
+  if (e->getBase()->getType()->isVectorType() &&
+      !isa<ExtVectorElementExpr>(e->getBase())) {
+    cgm.errorNYI(e->getSourceRange(), "emitArraySubscriptExpr: VectorType");
+    return LValue::makeAddr(Address::invalid(), e->getType(), LValueBaseInfo());
+  }
+
+  if (isa<ExtVectorElementExpr>(e->getBase())) {
+    cgm.errorNYI(e->getSourceRange(),
+                 "emitArraySubscriptExpr: ExtVectorElementExpr");
+    return LValue::makeAddr(Address::invalid(), e->getType(), LValueBaseInfo());
+  }
+
+  if (getContext().getAsVariableArrayType(e->getType())) {
+    cgm.errorNYI(e->getSourceRange(),
+                 "emitArraySubscriptExpr: VariableArrayType");
+    return LValue::makeAddr(Address::invalid(), e->getType(), LValueBaseInfo());
+  }
+
+  if (e->getType()->getAs<ObjCObjectType>()) {
+    cgm.errorNYI(e->getSourceRange(), "emitArraySubscriptExpr: ObjCObjectType");
+    return LValue::makeAddr(Address::invalid(), e->getType(), LValueBaseInfo());
+  }
+
+  // The index must always be an integer, which is not an aggregate.  Emit it
+  // in lexical order (this complexity is, sadly, required by C++17).
+  assert((e->getIdx() == e->getLHS() || e->getIdx() == e->getRHS()) &&
+         "index was neither LHS nor RHS");
+  const mlir::Value idx = emitScalarExpr(e->getIdx());
+  if (const Expr *array = getSimpleArrayDecayOperand(e->getBase())) {
+    LValue arrayLV;
+    if (const auto *ase = dyn_cast<ArraySubscriptExpr>(array))
+      arrayLV = emitArraySubscriptExpr(ase);
+    else
+      arrayLV = emitLValue(array);
+
+    // Propagate the alignment from the array itself to the result.
+    const Address addr = emitArraySubscriptPtr(
+        *this, cgm.getLoc(array->getBeginLoc()), cgm.getLoc(array->getEndLoc()),
+        arrayLV.getAddress(), e->getType(), idx, cgm.getLoc(e->getExprLoc()),
+        /*shouldDecay=*/true);
+
+    return LValue::makeAddr(addr, e->getType(), LValueBaseInfo());
+  }
+
+  // The base must be a pointer; emit it with an estimate of its alignment.
+  cgm.errorNYI(e->getSourceRange(),
+               "emitArraySubscriptExpr: The base must be a pointer");
+  return {};
+}
+
 LValue CIRGenFunction::emitBinaryOperatorLValue(const BinaryOperator *e) {
   // Comma expressions just emit their LHS then their RHS as an l-value.
   if (e->getOpcode() == BO_Comma) {

clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp

@@ -158,6 +158,16 @@ class ScalarExprEmitter : public StmtVisitor<ScalarExprEmitter, mlir::Value> {
   mlir::Value VisitCastExpr(CastExpr *e);
   mlir::Value VisitCallExpr(const CallExpr *e);
 
+  mlir::Value VisitArraySubscriptExpr(ArraySubscriptExpr *e) {
+    if (e->getBase()->getType()->isVectorType()) {
+      assert(!cir::MissingFeatures::scalableVectors());
+      cgf.getCIRGenModule().errorNYI("VisitArraySubscriptExpr: VectorType");
+      return {};
+    }
+    // Just load the lvalue formed by the subscript expression.
+    return emitLoadOfLValue(e);
+  }
+
   mlir::Value VisitExplicitCastExpr(ExplicitCastExpr *e) {
     return VisitCastExpr(e);
   }

clang/lib/CIR/CodeGen/CIRGenFunction.cpp

@@ -509,6 +509,8 @@ LValue CIRGenFunction::emitLValue(const Expr *e) {
                                std::string("l-value not implemented for '") +
                                    e->getStmtClassName() + "'");
     return LValue();
+  case Expr::ArraySubscriptExprClass:
+    return emitArraySubscriptExpr(cast<ArraySubscriptExpr>(e));
   case Expr::UnaryOperatorClass:
     return emitUnaryOpLValue(cast<UnaryOperator>(e));
   case Expr::BinaryOperatorClass:

clang/lib/CIR/CodeGen/CIRGenFunction.h

@@ -434,6 +434,8 @@ class CIRGenFunction : public CIRGenTypeCache {
   /// should be returned.
   RValue emitAnyExpr(const clang::Expr *e);
 
+  LValue emitArraySubscriptExpr(const clang::ArraySubscriptExpr *e);
+
   AutoVarEmission emitAutoVarAlloca(const clang::VarDecl &d);
 
   /// Emit code and set up symbol table for a variable declaration with auto,

clang/lib/CIR/CodeGen/CIRGenValue.h

@@ -23,6 +23,8 @@
 
 #include "mlir/IR/Value.h"
 
+#include "clang/CIR/MissingFeatures.h"
+
 namespace clang::CIRGen {
 
 /// This trivial value class is used to represent the result of an

clang/lib/CIR/CodeGen/CMakeLists.txt

@@ -8,6 +8,7 @@ get_property(dialect_libs GLOBAL PROPERTY MLIR_DIALECT_LIBS)
 
 add_clang_library(clangCIR
   CIRGenerator.cpp
+  CIRGenBuilder.cpp
   CIRGenCall.cpp
   CIRGenDecl.cpp
   CIRGenDeclOpenACC.cpp