Move the complex address expression out of DIVariable and into an extra

argument of the llvm.dbg.declare/llvm.dbg.value intrinsics.

Previously, DIVariable was a variable-length field that has an optional
reference to a Metadata array consisting of a variable number of
complex address expressions. In the case of OpPiece expressions this is
wasting a lot of storage in IR, because when an aggregate type is, e.g.,
SROA'd into all of its n individual members, the IR will contain n copies
of the DIVariable, all alike, only differing in the complex address
reference at the end.

By making the complex address into an extra argument of the
dbg.value/dbg.declare intrinsics, all of the pieces can reference the
same variable and the complex address expressions can be uniqued across
the CU, too.
Down the road, this will allow us to move other flags, such as
"indirection" out of the DIVariable, too.

The new intrinsics look like this:
declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr)
declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr)

This patch adds a new LLVM-local tag to DIExpressions, so we can detect
and pretty-print DIExpression metadata nodes.

What this patch doesn't do:

This patch does not touch the "Indirect" field in DIVariable; but moving
that into the expression would be a natural next step.

http://reviews.llvm.org/D4919
rdar://problem/17994491

Thanks to dblaikie and dexonsmith for reviewing this patch!

llvm-svn: 218778

3 files changed, 136 insertions, 91 deletions

diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index bbb8462652c..b4f327bab5c 100644
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -17,6 +17,7 @@
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DebugInfo.h"
 #include "llvm/IR/DiagnosticInfo.h"
+#include "llvm/IR/DIBuilder.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Instruction.h"
@@ -106,6 +107,20 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
     }
     break;
   }
+  case 'd': {
+    if (Name.startswith("dbg.declare") && F->arg_size() == 2) {
+      F->setName(Name + ".old");
+      NewFn = Intrinsic::getDeclaration(F->getParent(), Intrinsic::dbg_declare);
+      return true;
+    }
+    if (Name.startswith("dbg.value") && F->arg_size() == 3) {
+      F->setName(Name + ".old");
+      NewFn = Intrinsic::getDeclaration(F->getParent(), Intrinsic::dbg_value);
+      return true;
+    }
+    break;
+  }
+
   case 'o':
     // We only need to change the name to match the mangling including the
     // address space.
@@ -239,6 +254,22 @@ bool llvm::UpgradeGlobalVariable(GlobalVariable *GV) {
   return false;
 }
 
+static MDNode *getNodeField(const MDNode *DbgNode, unsigned Elt) {
+  if (!DbgNode || Elt >= DbgNode->getNumOperands())
+    return nullptr;
+  return dyn_cast_or_null<MDNode>(DbgNode->getOperand(Elt));
+}
+
+static DIExpression getExpression(Value *VarOperand, Function *F) {
+  // Old-style DIVariables have an optional expression as the 8th element.
+  DIExpression Expr(getNodeField(cast<MDNode>(VarOperand), 8));
+  if (!Expr) {
+    DIBuilder DIB(*F->getParent());
+    Expr = DIB.createExpression();
+  }
+  return Expr;
+}
+
 // UpgradeIntrinsicCall - Upgrade a call to an old intrinsic to be a call the
 // upgraded intrinsic. All argument and return casting must be provided in
 // order to seamlessly integrate with existing context.
@@ -402,12 +433,32 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
   }
 
   std::string Name = CI->getName().str();
-  CI->setName(Name + ".old");
+  if (!Name.empty())
+    CI->setName(Name + ".old");
 
   switch (NewFn->getIntrinsicID()) {
   default:
     llvm_unreachable("Unknown function for CallInst upgrade.");
 
+  // Upgrade debug intrinsics to use an additional DIExpression argument.
+  case Intrinsic::dbg_declare: {
+    auto NewCI =
+        Builder.CreateCall3(NewFn, CI->getArgOperand(0), CI->getArgOperand(1),
+                            getExpression(CI->getArgOperand(1), F), Name);
+    NewCI->setDebugLoc(CI->getDebugLoc());
+    CI->replaceAllUsesWith(NewCI);
+    CI->eraseFromParent();
+    return;
+  }
+  case Intrinsic::dbg_value: {
+    auto NewCI = Builder.CreateCall4(
+        NewFn, CI->getArgOperand(0), CI->getArgOperand(1), CI->getArgOperand(2),
+        getExpression(CI->getArgOperand(2), F), Name);
+    NewCI->setDebugLoc(CI->getDebugLoc());
+    CI->replaceAllUsesWith(NewCI);
+    CI->eraseFromParent();
+    return;
+  }
   case Intrinsic::ctlz:
   case Intrinsic::cttz:
     assert(CI->getNumArgOperands() == 1 &&
diff --git a/llvm/lib/IR/DIBuilder.cpp b/llvm/lib/IR/DIBuilder.cpp
index a5c09b60c4f..1ab52523d73 100644
--- a/llvm/lib/IR/DIBuilder.cpp
+++ b/llvm/lib/IR/DIBuilder.cpp
@@ -1042,50 +1042,28 @@ DIVariable DIBuilder::createLocalVariable(unsigned Tag, DIDescriptor Scope,
   return RetVar;
 }
 
-/// createComplexVariable - Create a new descriptor for the specified variable
-/// which has a complex address expression for its address.
-DIVariable DIBuilder::createComplexVariable(unsigned Tag, DIDescriptor Scope,
-                                            StringRef Name, DIFile F,
-                                            unsigned LineNo,
-                                            DITypeRef Ty,
-                                            ArrayRef<Value *> Addr,
-                                            unsigned ArgNo) {
-  assert(Addr.size() > 0 && "complex address is empty");
-  Value *Elts[] = {
-    GetTagConstant(VMContext, Tag),
-    getNonCompileUnitScope(Scope),
-    MDString::get(VMContext, Name),
-    F,
-    ConstantInt::get(Type::getInt32Ty(VMContext),
-                     (LineNo | (ArgNo << 24))),
-    Ty,
-    Constant::getNullValue(Type::getInt32Ty(VMContext)),
-    Constant::getNullValue(Type::getInt32Ty(VMContext)),
-    MDNode::get(VMContext, Addr)
-  };
-  return DIVariable(MDNode::get(VMContext, Elts));
+/// createExpression - Create a new descriptor for the specified
+/// variable which has a complex address expression for its address.
+/// @param Addr        An array of complex address operations.
+DIExpression DIBuilder::createExpression(ArrayRef<Value *> Addr) {
+  SmallVector<llvm::Value *, 16> Elts;
+  Elts.push_back(GetTagConstant(VMContext, DW_TAG_expression));
+  Elts.insert(Elts.end(), Addr.begin(), Addr.end());
+  return DIExpression(MDNode::get(VMContext, Elts));
 }
 
 /// createVariablePiece - Create a descriptor to describe one part
 /// of aggregate variable that is fragmented across multiple Values.
-DIVariable DIBuilder::createVariablePiece(DIVariable Variable,
-                                          unsigned OffsetInBytes,
-                                          unsigned SizeInBytes) {
+DIExpression DIBuilder::createPieceExpression(unsigned OffsetInBytes,
+                                              unsigned SizeInBytes) {
   assert(SizeInBytes > 0 && "zero-size piece");
   Value *Addr[] = {
-    ConstantInt::get(Type::getInt32Ty(VMContext), OpPiece),
-    ConstantInt::get(Type::getInt32Ty(VMContext), OffsetInBytes),
-    ConstantInt::get(Type::getInt32Ty(VMContext), SizeInBytes)
-  };
-
-  assert((Variable->getNumOperands() == 8 || Variable.isVariablePiece()) &&
-         "variable already has a complex address");
-  SmallVector<Value *, 9> Elts;
-  for (unsigned i = 0; i < 8; ++i)
-    Elts.push_back(Variable->getOperand(i));
+      GetTagConstant(VMContext, DW_TAG_expression),
+      ConstantInt::get(Type::getInt64Ty(VMContext), dwarf::DW_OP_piece),
+      ConstantInt::get(Type::getInt64Ty(VMContext), OffsetInBytes),
+      ConstantInt::get(Type::getInt64Ty(VMContext), SizeInBytes)};
 
-  Elts.push_back(MDNode::get(VMContext, Addr));
-  return DIVariable(MDNode::get(VMContext, Elts));
+  return DIExpression(MDNode::get(VMContext, Addr));
 }
 
 /// createFunction - Create a new descriptor for the specified function.
@@ -1292,6 +1270,7 @@ DILexicalBlock DIBuilder::createLexicalBlock(DIDescriptor Scope, DIFile File,
 
 /// insertDeclare - Insert a new llvm.dbg.declare intrinsic call.
 Instruction *DIBuilder::insertDeclare(Value *Storage, DIVariable VarInfo,
+                                      DIExpression Expr,
                                       Instruction *InsertBefore) {
   assert(Storage && "no storage passed to dbg.declare");
   assert(VarInfo.isVariable() &&
@@ -1299,12 +1278,13 @@ Instruction *DIBuilder::insertDeclare(Value *Storage, DIVariable VarInfo,
   if (!DeclareFn)
     DeclareFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_declare);
 
-  Value *Args[] = { MDNode::get(Storage->getContext(), Storage), VarInfo };
+  Value *Args[] = {MDNode::get(Storage->getContext(), Storage), VarInfo, Expr};
   return CallInst::Create(DeclareFn, Args, "", InsertBefore);
 }
 
 /// insertDeclare - Insert a new llvm.dbg.declare intrinsic call.
 Instruction *DIBuilder::insertDeclare(Value *Storage, DIVariable VarInfo,
+                                      DIExpression Expr,
                                       BasicBlock *InsertAtEnd) {
   assert(Storage && "no storage passed to dbg.declare");
   assert(VarInfo.isVariable() &&
@@ -1312,7 +1292,7 @@ Instruction *DIBuilder::insertDeclare(Value *Storage, DIVariable VarInfo,
   if (!DeclareFn)
     DeclareFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_declare);
 
-  Value *Args[] = { MDNode::get(Storage->getContext(), Storage), VarInfo };
+  Value *Args[] = {MDNode::get(Storage->getContext(), Storage), VarInfo, Expr};
 
   // If this block already has a terminator then insert this intrinsic
   // before the terminator.
@@ -1325,6 +1305,7 @@ Instruction *DIBuilder::insertDeclare(Value *Storage, DIVariable VarInfo,
 /// insertDbgValueIntrinsic - Insert a new llvm.dbg.value intrinsic call.
 Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
                                                 DIVariable VarInfo,
+                                                DIExpression Expr,
                                                 Instruction *InsertBefore) {
   assert(V && "no value passed to dbg.value");
   assert(VarInfo.isVariable() &&
@@ -1332,15 +1313,16 @@ Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
   if (!ValueFn)
     ValueFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_value);
 
-  Value *Args[] = { MDNode::get(V->getContext(), V),
-                    ConstantInt::get(Type::getInt64Ty(V->getContext()), Offset),
-                    VarInfo };
+  Value *Args[] = {MDNode::get(V->getContext(), V),
+                   ConstantInt::get(Type::getInt64Ty(V->getContext()), Offset),
+                   VarInfo, Expr};
   return CallInst::Create(ValueFn, Args, "", InsertBefore);
 }
 
 /// insertDbgValueIntrinsic - Insert a new llvm.dbg.value intrinsic call.
 Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
                                                 DIVariable VarInfo,
+                                                DIExpression Expr,
                                                 BasicBlock *InsertAtEnd) {
   assert(V && "no value passed to dbg.value");
   assert(VarInfo.isVariable() &&
@@ -1348,8 +1330,8 @@ Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
   if (!ValueFn)
     ValueFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_value);
 
-  Value *Args[] = { MDNode::get(V->getContext(), V),
-                    ConstantInt::get(Type::getInt64Ty(V->getContext()), Offset),
-                    VarInfo };
+  Value *Args[] = {MDNode::get(V->getContext(), V),
+                   ConstantInt::get(Type::getInt64Ty(V->getContext()), Offset),
+                   VarInfo, Expr};
   return CallInst::Create(ValueFn, Args, "", InsertAtEnd);
 }
diff --git a/llvm/lib/IR/DebugInfo.cpp b/llvm/lib/IR/DebugInfo.cpp
index 4274d82d158..4e36a8df3e1 100644
--- a/llvm/lib/IR/DebugInfo.cpp
+++ b/llvm/lib/IR/DebugInfo.cpp
@@ -49,7 +49,7 @@ bool DIDescriptor::Verify() const {
           DIObjCProperty(DbgNode).Verify() ||
           DITemplateTypeParameter(DbgNode).Verify() ||
           DITemplateValueParameter(DbgNode).Verify() ||
-          DIImportedEntity(DbgNode).Verify());
+          DIImportedEntity(DbgNode).Verify() || DIExpression(DbgNode).Verify());
 }
 
 static Value *getField(const MDNode *DbgNode, unsigned Elt) {
@@ -138,33 +138,9 @@ void DIDescriptor::replaceFunctionField(unsigned Elt, Function *F) {
   }
 }
 
-uint64_t DIVariable::getAddrElement(unsigned Idx) const {
-  DIDescriptor ComplexExpr = getDescriptorField(8);
-  if (Idx < ComplexExpr->getNumOperands())
-    if (auto *CI = dyn_cast_or_null<ConstantInt>(ComplexExpr->getOperand(Idx)))
-      return CI->getZExtValue();
-
-  assert(false && "non-existing complex address element requested");
-  return 0;
-}
-
 /// getInlinedAt - If this variable is inlined then return inline location.
 MDNode *DIVariable::getInlinedAt() const { return getNodeField(DbgNode, 7); }
 
-bool DIVariable::isVariablePiece() const {
-  return hasComplexAddress() && getAddrElement(0) == DIBuilder::OpPiece;
-}
-
-uint64_t DIVariable::getPieceOffset() const {
-  assert(isVariablePiece());
-  return getAddrElement(1);
-}
-
-uint64_t DIVariable::getPieceSize() const {
-  assert(isVariablePiece());
-  return getAddrElement(2);
-}
-
 /// Return the size reported by the variable's type.
 unsigned DIVariable::getSizeInBits(const DITypeIdentifierMap &Map) {
   DIType Ty = getType().resolve(Map);
@@ -178,8 +154,29 @@ unsigned DIVariable::getSizeInBits(const DITypeIdentifierMap &Map) {
   return Ty.getSizeInBits();
 }
 
+uint64_t DIExpression::getElement(unsigned Idx) const {
+  unsigned I = Idx + 1;
+  if (I < DbgNode->getNumOperands())
+    if (auto *CI = dyn_cast_or_null<ConstantInt>(DbgNode->getOperand(I)))
+      return CI->getZExtValue();
+
+  assert(false && "non-existing complex address element requested");
+  return 0;
+}
+
+bool DIExpression::isVariablePiece() const {
+  return getNumElements() && getElement(0) == dwarf::DW_OP_piece;
+}
 
+uint64_t DIExpression::getPieceOffset() const {
+  assert(isVariablePiece());
+  return getElement(1);
+}
 
+uint64_t DIExpression::getPieceSize() const {
+  assert(isVariablePiece());
+  return getElement(2);
+}
 
 //===----------------------------------------------------------------------===//
 // Predicates
@@ -357,6 +354,12 @@ bool DIDescriptor::isImportedEntity() const {
                      getTag() == dwarf::DW_TAG_imported_declaration);
 }
 
+/// \brief Return true if the specified tag is DW_TAG_imported_module or
+/// DW_TAG_imported_declaration.
+bool DIDescriptor::isExpression() const {
+  return DbgNode && (getTag() == dwarf::DW_TAG_expression);
+}
+
 //===----------------------------------------------------------------------===//
 // Simple Descriptor Constructors and other Methods
 //===----------------------------------------------------------------------===//
@@ -596,14 +599,16 @@ bool DIVariable::Verify() const {
   if (!fieldIsTypeRef(DbgNode, 5))
     return false;
 
-  // Variable without a complex expression.
-  if (DbgNode->getNumOperands() == 8)
+  // Variable without an inline location.
+  if (DbgNode->getNumOperands() == 7)
     return true;
 
-  // Make sure the complex expression is an MDNode.
-  return (DbgNode->getNumOperands() == 9 && fieldIsMDNode(DbgNode, 8));
+  return DbgNode->getNumOperands() == 8;
 }
 
+/// Verify - Verify that a variable descriptor is well formed.
+bool DIExpression::Verify() const { return isExpression(); }
+
 /// Verify - Verify that a location descriptor is well formed.
 bool DILocation::Verify() const {
   if (!DbgNode)
@@ -936,19 +941,6 @@ DIVariable llvm::cleanseInlinedVariable(MDNode *DV, LLVMContext &VMContext) {
   return DIVariable(MDNode::get(VMContext, Elts));
 }
 
-
-/// getEntireVariable - Remove OpPiece exprs from the variable.
-DIVariable llvm::getEntireVariable(DIVariable DV) {
-  if (!DV.isVariablePiece())
-    return DV;
-
-  SmallVector<Value *, 8> Elts;
-  for (unsigned i = 0; i < 8; ++i)
-    Elts.push_back(DV->getOperand(i));
-
-  return DIVariable(MDNode::get(DV->getContext(), Elts));
-}
-
 /// getDISubprogram - Find subprogram that is enclosing this scope.
 DISubprogram llvm::getDISubprogram(const MDNode *Scope) {
   DIDescriptor D(Scope);
@@ -1293,6 +1285,8 @@ void DIDescriptor::print(raw_ostream &OS) const {
     DINameSpace(DbgNode).printInternal(OS);
   } else if (this->isScope()) {
     DIScope(DbgNode).printInternal(OS);
+  } else if (this->isExpression()) {
+    DIExpression(DbgNode).printInternal(OS);
   }
 }
 
@@ -1442,10 +1436,28 @@ void DIVariable::printInternal(raw_ostream &OS) const {
     OS << " [" << Res << ']';
 
   OS << " [line " << getLineNumber() << ']';
+}
 
-  if (isVariablePiece())
-    OS << " [piece, size " << getPieceSize()
-       << ", offset " << getPieceOffset() << ']';
+void DIExpression::printInternal(raw_ostream &OS) const {
+  for (unsigned I = 0; I < getNumElements(); ++I) {
+    uint64_t OpCode = getElement(I);
+    OS << " [" << OperationEncodingString(OpCode);
+    switch (OpCode) {
+    case DW_OP_plus: {
+      OS << " " << getElement(++I);
+      break;
+    }
+    case DW_OP_piece: {
+      unsigned Offset = getElement(++I);
+      unsigned Size = getElement(++I);
+      OS << " offset=" << Offset << ", size= " << Size;
+      break;
+    }
+    default:
+      break;
+    }
+    OS << "]";
+  }
 }
 
 void DIObjCProperty::printInternal(raw_ostream &OS) const {