[flang] Handle volatility in lowering and codegen (llvm#135311)

* Enable lowering and conversion patterns to pass volatility information from higher level operations to lower level ones. * Enable codegen to pass volatility to LLVM dialect ops by setting an attribute on loads, stores, and memory intrinsics. * Add utilities for passing along the volatility from an input type to an output type. To introduce volatile types into the IR, entities with the volatile attribute will be given a volatile type in the bridge; this is not enabled in this patch. User code should not result in IR with volatile types yet, so this patch contains no tests with Fortran source, only IR that already contains volatile types. Part 3 of llvm#132486.

Author: ashermancinelli

flang/include/flang/Optimizer/Builder/BoxValue.h

@@ -236,7 +236,7 @@ class AbstractIrBox : public AbstractBox, public AbstractArrayBox {
  auto ty = getBoxTy().getEleTy();
  if (fir::isa_ref_type(ty))
  return ty;
- return fir::ReferenceType::get(ty);
+ return fir::ReferenceType::get(ty, fir::isa_volatile_type(ty));
  }
 
  /// Get the scalar type related to the described entity

flang/include/flang/Optimizer/Builder/FIRBuilder.h

@@ -402,6 +402,10 @@ class FirOpBuilder : public mlir::OpBuilder, public mlir::OpBuilder::Listener {
  mlir::Value createConvertWithVolatileCast(mlir::Location loc, mlir::Type toTy,
  mlir::Value val);
 
+ /// Cast \p value to have \p isVolatile volatility.
+ mlir::Value createVolatileCast(mlir::Location loc, bool isVolatile,
+ mlir::Value value);
+
  /// Create a fir.store of \p val into \p addr. A lazy conversion
  /// of \p val to the element type of \p addr is created if needed.
  void createStoreWithConvert(mlir::Location loc, mlir::Value val,

flang/lib/Lower/ConvertCall.cpp

@@ -327,7 +327,6 @@ Fortran::lower::genCallOpAndResult(
  charFuncPointerLength = charBox->getLen();
  }
  }
-
  const bool isExprCall =
  converter.getLoweringOptions().getLowerToHighLevelFIR() &&
  callSiteType.getNumResults() == 1 &&
@@ -519,7 +518,9 @@ Fortran::lower::genCallOpAndResult(
  // Do not attempt any reboxing here that could break this.
  bool legacyLowering =
  !converter.getLoweringOptions().getLowerToHighLevelFIR();
- cast = builder.convertWithSemantics(loc, snd, fst,
+ bool isVolatile = fir::isa_volatile_type(snd);
+ cast = builder.createVolatileCast(loc, isVolatile, fst);
+ cast = builder.convertWithSemantics(loc, snd, cast,
  callingImplicitInterface,
  /*allowRebox=*/legacyLowering);
  }
@@ -1415,6 +1416,11 @@ static PreparedDummyArgument preparePresentUserCallActualArgument(
  addr = hlfir::genVariableRawAddress(loc, builder, entity);
  }
 
+ // If the volatility of the input type does not match the dummy type,
+ // we need to cast the argument.
+ const bool isToTypeVolatile = fir::isa_volatile_type(dummyTypeWithActualRank);
+ addr = builder.createVolatileCast(loc, isToTypeVolatile, addr);
+
  // For ranked actual passed to assumed-rank dummy, the cast to assumed-rank
  // box is inserted when building the fir.call op. Inserting it here would
  // cause the fir.if results to be assumed-rank in case of OPTIONAL dummy,

flang/lib/Lower/ConvertExprToHLFIR.cpp

@@ -212,19 +212,25 @@ class HlfirDesignatorBuilder {
  auto charType = mlir::dyn_cast<fir::CharacterType>(resultValueType);
  if (charType && charType.hasDynamicLen())
  return fir::BoxCharType::get(charType.getContext(), charType.getFKind());
+
+ // When volatile is enabled, enable volatility on the designatory type.
+ const bool isVolatile = false;
+
  // Arrays with non default lower bounds or dynamic length or dynamic extent
  // need a fir.box to hold the dynamic or lower bound information.
  if (fir::hasDynamicSize(resultValueType) ||
  mayHaveNonDefaultLowerBounds(partInfo))
- return fir::BoxType::get(resultValueType);
+ return fir::BoxType::get(resultValueType, isVolatile);
+
  // Non simply contiguous ref require a fir.box to carry the byte stride.
  if (mlir::isa<fir::SequenceType>(resultValueType) &&
  !Fortran::evaluate::IsSimplyContiguous(
  designatorNode, getConverter().getFoldingContext(),
  /*namedConstantSectionsAreAlwaysContiguous=*/false))
- return fir::BoxType::get(resultValueType);
+ return fir::BoxType::get(resultValueType, isVolatile);
+
  // Other designators can be handled as raw addresses.
- return fir::ReferenceType::get(resultValueType);
+ return fir::ReferenceType::get(resultValueType, isVolatile);
  }
 
  template <typename T>
@@ -1824,7 +1830,10 @@ class HlfirBuilder {
  assert(compType && "failed to retrieve component type");
  mlir::Value compShape =
  designatorBuilder.genComponentShape(sym, compType);
- mlir::Type designatorType = builder.getRefType(compType);
+ const bool isDesignatorVolatile =
+ fir::isa_volatile_type(baseOp.getType());
+ mlir::Type designatorType =
+ builder.getRefType(compType, isDesignatorVolatile);
 
  mlir::Type fieldElemType = hlfir::getFortranElementType(compType);
  llvm::SmallVector<mlir::Value, 1> typeParams;

flang/lib/Lower/HostAssociations.cpp

@@ -194,8 +194,8 @@ class CapturedProcedure : public CapturedSymbols<CapturedProcedure> {
  fir::FirOpBuilder &builder = converter.getFirOpBuilder();
  mlir::Type typeInTuple = fir::dyn_cast_ptrEleTy(args.addrInTuple.getType());
  assert(typeInTuple && "addrInTuple must be an address");
- mlir::Value castBox = builder.createConvert(args.loc, typeInTuple,
-   fir::getBase(args.hostValue));
+ mlir::Value castBox = builder.createConvertWithVolatileCast(
+ args.loc, typeInTuple, fir::getBase(args.hostValue));
  builder.create<fir::StoreOp>(args.loc, castBox, args.addrInTuple);
  }
 
@@ -265,8 +265,8 @@ class CapturedPolymorphicScalar
  mlir::Location loc = args.loc;
  mlir::Type typeInTuple = fir::dyn_cast_ptrEleTy(args.addrInTuple.getType());
  assert(typeInTuple && "addrInTuple must be an address");
- mlir::Value castBox = builder.createConvert(args.loc, typeInTuple,
-   fir::getBase(args.hostValue));
+ mlir::Value castBox = builder.createConvertWithVolatileCast(
+ args.loc, typeInTuple, fir::getBase(args.hostValue));
  if (Fortran::semantics::IsOptional(sym)) {
  auto isPresent =
  builder.create<fir::IsPresentOp>(loc, builder.getI1Type(), castBox);
@@ -329,8 +329,8 @@ class CapturedAllocatableAndPointer
  fir::FirOpBuilder &builder = converter.getFirOpBuilder();
  mlir::Type typeInTuple = fir::dyn_cast_ptrEleTy(args.addrInTuple.getType());
  assert(typeInTuple && "addrInTuple must be an address");
- mlir::Value castBox = builder.createConvert(args.loc, typeInTuple,
-   fir::getBase(args.hostValue));
+ mlir::Value castBox = builder.createConvertWithVolatileCast(
+ args.loc, typeInTuple, fir::getBase(args.hostValue));
  builder.create<fir::StoreOp>(args.loc, castBox, args.addrInTuple);
  }
  static void getFromTuple(const GetFromTuple &args,

flang/lib/Lower/IO.cpp

@@ -720,7 +720,8 @@ static void genOutputItemList(
  fir::factory::CharacterExprHelper helper{builder, loc};
  if (mlir::isa<fir::BoxType>(argType)) {
  mlir::Value box = fir::getBase(converter.genExprBox(loc, *expr, stmtCtx));
- outputFuncArgs.push_back(builder.createConvert(loc, argType, box));
+ outputFuncArgs.push_back(
+ builder.createConvertWithVolatileCast(loc, argType, box));
  if (containsDerivedType(itemTy))
  outputFuncArgs.push_back(getNonTbpDefinedIoTableAddr(converter));
  } else if (helper.isCharacterScalar(itemTy)) {
@@ -730,9 +731,9 @@ static void genOutputItemList(
  if (!exv.getCharBox())
  llvm::report_fatal_error(
  "internal error: scalar character not in CharBox");
- outputFuncArgs.push_back(builder.createConvert(
+ outputFuncArgs.push_back(builder.createConvertWithVolatileCast(
  loc, outputFunc.getFunctionType().getInput(1), fir::getBase(exv)));
- outputFuncArgs.push_back(builder.createConvert(
+ outputFuncArgs.push_back(builder.createConvertWithVolatileCast(
  loc, outputFunc.getFunctionType().getInput(2), fir::getLen(exv)));
  } else {
  fir::ExtendedValue itemBox = converter.genExprValue(loc, expr, stmtCtx);
@@ -743,7 +744,8 @@ static void genOutputItemList(
  outputFuncArgs.push_back(parts.first);
  outputFuncArgs.push_back(parts.second);
  } else {
- itemValue = builder.createConvert(loc, argType, itemValue);
+ itemValue =
+ builder.createConvertWithVolatileCast(loc, argType, itemValue);
  outputFuncArgs.push_back(itemValue);
  }
  }
@@ -827,7 +829,8 @@ createIoRuntimeCallForItem(Fortran::lower::AbstractConverter &converter,
  mlir::Value box = fir::getBase(item);
  auto boxTy = mlir::dyn_cast<fir::BaseBoxType>(box.getType());
  assert(boxTy && "must be previously emboxed");
- inputFuncArgs.push_back(builder.createConvert(loc, argType, box));
+ auto casted = builder.createConvertWithVolatileCast(loc, argType, box);
+ inputFuncArgs.push_back(casted);
  if (containsDerivedType(boxTy))
  inputFuncArgs.push_back(getNonTbpDefinedIoTableAddr(converter));
  } else {

flang/lib/Optimizer/Builder/FIRBuilder.cpp

@@ -577,14 +577,20 @@ mlir::Value fir::FirOpBuilder::convertWithSemantics(
  return createConvert(loc, toTy, val);
 }
 
+mlir::Value fir::FirOpBuilder::createVolatileCast(mlir::Location loc,
+ bool isVolatile,
+ mlir::Value val) {
+ mlir::Type volatileAdjustedType =
+ fir::updateTypeWithVolatility(val.getType(), isVolatile);
+ if (volatileAdjustedType == val.getType())
+ return val;
+ return create<fir::VolatileCastOp>(loc, volatileAdjustedType, val);
+}
+
 mlir::Value fir::FirOpBuilder::createConvertWithVolatileCast(mlir::Location loc,
  mlir::Type toTy,
  mlir::Value val) {
- if (fir::isa_volatile_type(val.getType()) != fir::isa_volatile_type(toTy)) {
- mlir::Type volatileAdjustedType = fir::updateTypeWithVolatility(
- val.getType(), fir::isa_volatile_type(toTy));
- val = create<fir::VolatileCastOp>(loc, volatileAdjustedType, val);
- }
+ val = createVolatileCast(loc, fir::isa_volatile_type(toTy), val);
  return createConvert(loc, toTy, val);
 }
 

flang/lib/Optimizer/Builder/HLFIRTools.cpp

@@ -382,7 +382,9 @@ hlfir::Entity hlfir::genVariableBox(mlir::Location loc,
  mlir::Value addr = var.getBase();
  if (mlir::isa<fir::BoxCharType>(var.getType()))
  addr = genVariableRawAddress(loc, builder, var);
- mlir::Type boxType = fir::BoxType::get(var.getElementOrSequenceType());
+ const bool isVolatile = fir::isa_volatile_type(var.getType());
+ mlir::Type boxType =
+ fir::BoxType::get(var.getElementOrSequenceType(), isVolatile);
  if (forceBoxType) {
  boxType = forceBoxType;
  mlir::Type baseType =
@@ -793,15 +795,16 @@ mlir::Type hlfir::getVariableElementType(hlfir::Entity variable) {
  if (variable.isScalar())
  return variable.getType();
  mlir::Type eleTy = variable.getFortranElementType();
+ const bool isVolatile = fir::isa_volatile_type(variable.getType());
  if (variable.isPolymorphic())
- return fir::ClassType::get(eleTy);
+ return fir::ClassType::get(eleTy, isVolatile);
  if (auto charType = mlir::dyn_cast<fir::CharacterType>(eleTy)) {
  if (charType.hasDynamicLen())
  return fir::BoxCharType::get(charType.getContext(), charType.getFKind());
  } else if (fir::isRecordWithTypeParameters(eleTy)) {
- return fir::BoxType::get(eleTy);
+ return fir::BoxType::get(eleTy, isVolatile);
  }
- return fir::ReferenceType::get(eleTy);
+ return fir::ReferenceType::get(eleTy, isVolatile);
 }
 
 mlir::Type hlfir::getEntityElementType(hlfir::Entity entity) {

flang/lib/Optimizer/Builder/MutableBox.cpp

@@ -76,7 +76,9 @@ createNewFirBox(fir::FirOpBuilder &builder, mlir::Location loc,
  cleanedLengths = lengths;
  }
  mlir::Value emptySlice;
- return builder.create<fir::EmboxOp>(loc, box.getBoxTy(), cleanedAddr, shape,
+ auto boxType = fir::updateTypeWithVolatility(
+ box.getBoxTy(), fir::isa_volatile_type(cleanedAddr.getType()));
+ return builder.create<fir::EmboxOp>(loc, boxType, cleanedAddr, shape,
  emptySlice, cleanedLengths, tdesc);
 }
 
@@ -281,6 +283,9 @@ class MutablePropertyWriter {
  unsigned allocator = kDefaultAllocator) {
  mlir::Value irBox = createNewFirBox(builder, loc, box, addr, lbounds,
  extents, lengths, tdesc);
+ const bool valueTypeIsVolatile =
+ fir::isa_volatile_type(fir::unwrapRefType(box.getAddr().getType()));
+ irBox = builder.createVolatileCast(loc, valueTypeIsVolatile, irBox);
  builder.create<fir::StoreOp>(loc, irBox, box.getAddr());
  }
 
@@ -346,7 +351,8 @@ mlir::Value fir::factory::createUnallocatedBox(
  baseBoxType = baseBoxType.getBoxTypeWithNewShape(/*rank=*/0);
  auto baseAddrType = baseBoxType.getEleTy();
  if (!fir::isa_ref_type(baseAddrType))
- baseAddrType = builder.getRefType(baseAddrType);
+ baseAddrType =
+ builder.getRefType(baseAddrType, fir::isa_volatile_type(baseBoxType));
  auto type = fir::unwrapRefType(baseAddrType);
  auto eleTy = fir::unwrapSequenceType(type);
  if (auto recTy = mlir::dyn_cast<fir::RecordType>(eleTy))
@@ -516,7 +522,7 @@ void fir::factory::associateMutableBox(fir::FirOpBuilder &builder,
  source.match(
  [&](const fir::PolymorphicValue &p) {
  mlir::Value sourceBox;
- if (auto polyBox = source.getBoxOf<fir::PolymorphicValue>())
+ if (auto *polyBox = source.getBoxOf<fir::PolymorphicValue>())
  sourceBox = polyBox->getSourceBox();
  writer.updateMutableBox(p.getAddr(), /*lbounds=*/std::nullopt,
  /*extents=*/std::nullopt,

flang/lib/Optimizer/CodeGen/CodeGen.cpp

@@ -689,6 +689,20 @@ struct CmpcOpConversion : public fir::FIROpConversion<fir::CmpcOp> {
  }
 };
 
+/// fir.volatile_cast is only useful at the fir level. Once we lower to LLVM,
+/// volatility is described by setting volatile attributes on the LLVM ops.
+struct VolatileCastOpConversion
+ : public fir::FIROpConversion<fir::VolatileCastOp> {
+ using FIROpConversion::FIROpConversion;
+
+ llvm::LogicalResult
+ matchAndRewrite(fir::VolatileCastOp volatileCast, OpAdaptor adaptor,
+ mlir::ConversionPatternRewriter &rewriter) const override {
+ rewriter.replaceOp(volatileCast, adaptor.getOperands()[0]);
+ return mlir::success();
+ }
+};
+
 /// convert value of from-type to value of to-type
 struct ConvertOpConversion : public fir::FIROpConversion<fir::ConvertOp> {
  using FIROpConversion::FIROpConversion;
@@ -3224,6 +3238,7 @@ struct LoadOpConversion : public fir::FIROpConversion<fir::LoadOp> {
  mlir::ConversionPatternRewriter &rewriter) const override {
 
  mlir::Type llvmLoadTy = convertObjectType(load.getType());
+ const bool isVolatile = fir::isa_volatile_type(load.getMemref().getType());
  if (auto boxTy = mlir::dyn_cast<fir::BaseBoxType>(load.getType())) {
  // fir.box is a special case because it is considered an ssa value in
  // fir, but it is lowered as a pointer to a descriptor. So
@@ -3253,16 +3268,17 @@ struct LoadOpConversion : public fir::FIROpConversion<fir::LoadOp> {
  mlir::Value boxSize =
  computeBoxSize(loc, boxTypePair, inputBoxStorage, rewriter);
  auto memcpy = rewriter.create<mlir::LLVM::MemcpyOp>(
- loc, newBoxStorage, inputBoxStorage, boxSize, /*isVolatile=*/false);
+ loc, newBoxStorage, inputBoxStorage, boxSize, isVolatile);
 
  if (std::optional<mlir::ArrayAttr> optionalTag = load.getTbaa())
  memcpy.setTBAATags(*optionalTag);
  else
  attachTBAATag(memcpy, boxTy, boxTy, nullptr);
  rewriter.replaceOp(load, newBoxStorage);
  } else {
- auto loadOp = rewriter.create<mlir::LLVM::LoadOp>(
+ mlir::LLVM::LoadOp loadOp = rewriter.create<mlir::LLVM::LoadOp>(
  load.getLoc(), llvmLoadTy, adaptor.getOperands(), load->getAttrs());
+ loadOp.setVolatile_(isVolatile);
  if (std::optional<mlir::ArrayAttr> optionalTag = load.getTbaa())
  loadOp.setTBAATags(*optionalTag);
  else
@@ -3540,17 +3556,22 @@ struct StoreOpConversion : public fir::FIROpConversion<fir::StoreOp> {
  mlir::Value llvmValue = adaptor.getValue();
  mlir::Value llvmMemref = adaptor.getMemref();
  mlir::LLVM::AliasAnalysisOpInterface newOp;
+ const bool isVolatile = fir::isa_volatile_type(store.getMemref().getType());
  if (auto boxTy = mlir::dyn_cast<fir::BaseBoxType>(storeTy)) {
  mlir::Type llvmBoxTy = lowerTy().convertBoxTypeAsStruct(boxTy);
  // Always use memcpy because LLVM is not as effective at optimizing
  // aggregate loads/stores as it is optimizing memcpy.
  TypePair boxTypePair{boxTy, llvmBoxTy};
  mlir::Value boxSize =
  computeBoxSize(loc, boxTypePair, llvmValue, rewriter);
- newOp = rewriter.create<mlir::LLVM::MemcpyOp>(
- loc, llvmMemref, llvmValue, boxSize, /*isVolatile=*/false);
+ newOp = rewriter.create<mlir::LLVM::MemcpyOp>(loc, llvmMemref, llvmValue,
+  boxSize, isVolatile);
  } else {
- newOp = rewriter.create<mlir::LLVM::StoreOp>(loc, llvmValue, llvmMemref);
+ mlir::LLVM::StoreOp storeOp =
+ rewriter.create<mlir::LLVM::StoreOp>(loc, llvmValue, llvmMemref);
+ if (isVolatile)
+ storeOp.setVolatile_(true);
+ newOp = storeOp;
  }
  if (std::optional<mlir::ArrayAttr> optionalTag = store.getTbaa())
  newOp.setTBAATags(*optionalTag);
@@ -4193,21 +4214,22 @@ void fir::populateFIRToLLVMConversionPatterns(
  BoxIsAllocOpConversion, BoxIsArrayOpConversion, BoxIsPtrOpConversion,
  BoxOffsetOpConversion, BoxProcHostOpConversion, BoxRankOpConversion,
  BoxTypeCodeOpConversion, BoxTypeDescOpConversion, CallOpConversion,
- CmpcOpConversion, ConvertOpConversion, CoordinateOpConversion,
- CopyOpConversion, DTEntryOpConversion, DeclareOpConversion,
- DivcOpConversion, EmboxOpConversion, EmboxCharOpConversion,
- EmboxProcOpConversion, ExtractValueOpConversion, FieldIndexOpConversion,
- FirEndOpConversion, FreeMemOpConversion, GlobalLenOpConversion,
- GlobalOpConversion, InsertOnRangeOpConversion, IsPresentOpConversion,
- LenParamIndexOpConversion, LoadOpConversion, MulcOpConversion,
- NegcOpConversion, NoReassocOpConversion, SelectCaseOpConversion,
- SelectOpConversion, SelectRankOpConversion, SelectTypeOpConversion,
- ShapeOpConversion, ShapeShiftOpConversion, ShiftOpConversion,
- SliceOpConversion, StoreOpConversion, StringLitOpConversion,
- SubcOpConversion, TypeDescOpConversion, TypeInfoOpConversion,
- UnboxCharOpConversion, UnboxProcOpConversion, UndefOpConversion,
- UnreachableOpConversion, XArrayCoorOpConversion, XEmboxOpConversion,
- XReboxOpConversion, ZeroOpConversion>(converter, options);
+ CmpcOpConversion, VolatileCastOpConversion, ConvertOpConversion,
+ CoordinateOpConversion, CopyOpConversion, DTEntryOpConversion,
+ DeclareOpConversion, DivcOpConversion, EmboxOpConversion,
+ EmboxCharOpConversion, EmboxProcOpConversion, ExtractValueOpConversion,
+ FieldIndexOpConversion, FirEndOpConversion, FreeMemOpConversion,
+ GlobalLenOpConversion, GlobalOpConversion, InsertOnRangeOpConversion,
+ IsPresentOpConversion, LenParamIndexOpConversion, LoadOpConversion,
+ MulcOpConversion, NegcOpConversion, NoReassocOpConversion,
+ SelectCaseOpConversion, SelectOpConversion, SelectRankOpConversion,
+ SelectTypeOpConversion, ShapeOpConversion, ShapeShiftOpConversion,
+ ShiftOpConversion, SliceOpConversion, StoreOpConversion,
+ StringLitOpConversion, SubcOpConversion, TypeDescOpConversion,
+ TypeInfoOpConversion, UnboxCharOpConversion, UnboxProcOpConversion,
+ UndefOpConversion, UnreachableOpConversion, XArrayCoorOpConversion,
+ XEmboxOpConversion, XReboxOpConversion, ZeroOpConversion>(converter,
+ options);
 
  // Patterns that are populated without a type converter do not trigger
  // target materializations for the operands of the root op.