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!

Note: I accidentally committed a bogus older version of this patch previously.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/CodeGen/MachineInstr.h

@@ -244,12 +244,22 @@ public:
   ///
   DebugLoc getDebugLoc() const { return debugLoc; }
 
-  /// getDebugVariable() - Return the debug variable referenced by
+  /// \brief Return the debug variable referenced by
   /// this DBG_VALUE instruction.
   DIVariable getDebugVariable() const {
     assert(isDebugValue() && "not a DBG_VALUE");
-    const MDNode *Var = getOperand(getNumOperands() - 1).getMetadata();
-    return DIVariable(Var);
+    DIVariable Var(getOperand(2).getMetadata());
+    assert(Var.Verify() && "not a DIVariable");
+    return Var;
+  }
+
+  /// \brief Return the complex address expression referenced by
+  /// this DBG_VALUE instruction.
+  DIExpression getDebugExpression() const {
+    assert(isDebugValue() && "not a DBG_VALUE");
+    DIExpression Expr(getOperand(3).getMetadata());
+    assert(Expr.Verify() && "not a DIExpression");
+    return Expr;
   }
 
   /// emitError - Emit an error referring to the source location of this

include/llvm/CodeGen/MachineInstrBuilder.h

@@ -170,6 +170,8 @@ public:
 
   const MachineInstrBuilder &addMetadata(const MDNode *MD) const {
     MI->addOperand(*MF, MachineOperand::CreateMetadata(MD));
+    assert((MI->isDebugValue() ? MI->getDebugVariable().Verify() : true) &&
+           "first MDNode argument of a DBG_VALUE not a DIVariable");
     return *this;
   }
 
@@ -345,24 +347,25 @@ inline MachineInstrBuilder BuildMI(MachineBasicBlock *BB,
 /// address.  The convention is that a DBG_VALUE is indirect iff the
 /// second operand is an immediate.
 ///
-inline MachineInstrBuilder BuildMI(MachineFunction &MF,
-                                   DebugLoc DL,
-                                   const MCInstrDesc &MCID,
-                                   bool IsIndirect,
-                                   unsigned Reg,
-                                   unsigned Offset,
-                                   const MDNode *MD) {
+inline MachineInstrBuilder BuildMI(MachineFunction &MF, DebugLoc DL,
+                                   const MCInstrDesc &MCID, bool IsIndirect,
+                                   unsigned Reg, unsigned Offset,
+                                   const MDNode *Variable, const MDNode *Expr) {
+  assert(DIVariable(Variable).Verify() && "not a DIVariable");
+  assert(DIExpression(Expr).Verify() && "not a DIExpression");
   if (IsIndirect)
     return BuildMI(MF, DL, MCID)
-      .addReg(Reg, RegState::Debug)
-      .addImm(Offset)
-      .addMetadata(MD);
+        .addReg(Reg, RegState::Debug)
+        .addImm(Offset)
+        .addMetadata(Variable)
+        .addMetadata(Expr);
   else {
     assert(Offset == 0 && "A direct address cannot have an offset.");
     return BuildMI(MF, DL, MCID)
-      .addReg(Reg, RegState::Debug)
-      .addReg(0U, RegState::Debug)
-      .addMetadata(MD);
+        .addReg(Reg, RegState::Debug)
+        .addReg(0U, RegState::Debug)
+        .addMetadata(Variable)
+        .addMetadata(Expr);
   }
 }
 
@@ -371,15 +374,15 @@ inline MachineInstrBuilder BuildMI(MachineFunction &MF,
 /// address and inserts it at position I.
 ///
 inline MachineInstrBuilder BuildMI(MachineBasicBlock &BB,
-                                   MachineBasicBlock::iterator I,
-                                   DebugLoc DL,
-                                   const MCInstrDesc &MCID,
-                                   bool IsIndirect,
-                                   unsigned Reg,
-                                   unsigned Offset,
-                                   const MDNode *MD) {
+                                   MachineBasicBlock::iterator I, DebugLoc DL,
+                                   const MCInstrDesc &MCID, bool IsIndirect,
+                                   unsigned Reg, unsigned Offset,
+                                   const MDNode *Variable, const MDNode *Expr) {
+  assert(DIVariable(Variable).Verify() && "not a DIVariable");
+  assert(DIExpression(Expr).Verify() && "not a DIExpression");
   MachineFunction &MF = *BB.getParent();
-  MachineInstr *MI = BuildMI(MF, DL, MCID, IsIndirect, Reg, Offset, MD);
+  MachineInstr *MI =
+      BuildMI(MF, DL, MCID, IsIndirect, Reg, Offset, Variable, Expr);
   BB.insert(I, MI);
   return MachineInstrBuilder(MF, MI);
 }

include/llvm/CodeGen/MachineModuleInfo.h

@@ -166,6 +166,7 @@ public:
 
   struct VariableDbgInfo {
     TrackingVH<MDNode> Var;
+    TrackingVH<MDNode> Expr;
     unsigned Slot;
     DebugLoc Loc;
   };
@@ -390,8 +391,9 @@ public:
 
   /// setVariableDbgInfo - Collect information used to emit debugging
   /// information of a variable.
-  void setVariableDbgInfo(MDNode *N, unsigned Slot, DebugLoc Loc) {
-    VariableDbgInfo Info = { N, Slot, Loc };
+  void setVariableDbgInfo(MDNode *Var, MDNode *Expr, unsigned Slot,
+                          DebugLoc Loc) {
+    VariableDbgInfo Info = {Var, Expr, Slot, Loc};
     VariableDbgInfos.push_back(std::move(Info));
   }
 

include/llvm/CodeGen/SelectionDAG.h

@@ -984,15 +984,18 @@ public:
 
   /// getDbgValue - Creates a SDDbgValue node.
   ///
-  SDDbgValue *getDbgValue(MDNode *MDPtr, SDNode *N, unsigned R,
-			  bool IsIndirect, uint64_t Off,
-                          DebugLoc DL, unsigned O);
-  /// Constant.
-  SDDbgValue *getConstantDbgValue(MDNode *MDPtr, const Value *C, uint64_t Off,
-				  DebugLoc DL, unsigned O);
-  /// Frame index.
-  SDDbgValue *getFrameIndexDbgValue(MDNode *MDPtr, unsigned FI, uint64_t Off,
-				    DebugLoc DL, unsigned O);
+  /// SDNode
+  SDDbgValue *getDbgValue(MDNode *Var, MDNode *Expr, SDNode *N, unsigned R,
+                          bool IsIndirect, uint64_t Off, DebugLoc DL,
+                          unsigned O);
+
+  /// Constant
+  SDDbgValue *getConstantDbgValue(MDNode *Var, MDNode *Expr, const Value *C,
+                                  uint64_t Off, DebugLoc DL, unsigned O);
+
+  /// FrameIndex
+  SDDbgValue *getFrameIndexDbgValue(MDNode *Var, MDNode *Expr, unsigned FI,
+                                    uint64_t Off, DebugLoc DL, unsigned O);
 
   /// RemoveDeadNode - Remove the specified node from the system. If any of its
   /// operands then becomes dead, remove them as well. Inform UpdateListener