Add a new optimization pass: Stack Coloring, that merges disjoint static allocations (allocas). Allocas are known to be

disjoint if they are marked by disjoint lifetime markers (@llvm.lifetime.XXX intrinsics).



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163299 91177308-0d34-0410-b5e6-96231b3b80d8

docs/ReleaseNotes.html

@@ -499,6 +499,13 @@ Release Notes</a>.</h1>
 
 <div>
 
+<p>Stack Coloring - We have implemented a new optimization pass
+  to merge stack objects which are used in disjoin areas of the code.
+  This optimization reduces the required stack space significantly, in cases
+  where it is clear to the optimizer that the stack slot is not shared.
+  We use the lifetime markers to tell the codegen that a certain alloca
+  is used within a region.</p>
+
 <p>We have put a significant amount of work into the code generator
    infrastructure, which allows us to implement more aggressive algorithms and
    make it run faster:</p>

include/llvm/ADT/BitVector.h

@@ -311,7 +311,7 @@ public:
     return !(*this == RHS);
   }
 
-  // Intersection, union, disjoint union.
+  /// Intersection, union, disjoint union.
   BitVector &operator&=(const BitVector &RHS) {
     unsigned ThisWords = NumBitWords(size());
     unsigned RHSWords  = NumBitWords(RHS.size());
@@ -328,7 +328,7 @@ public:
     return *this;
   }
 
-  // reset - Reset bits that are set in RHS. Same as *this &= ~RHS.
+  /// reset - Reset bits that are set in RHS. Same as *this &= ~RHS.
   BitVector &reset(const BitVector &RHS) {
     unsigned ThisWords = NumBitWords(size());
     unsigned RHSWords  = NumBitWords(RHS.size());
@@ -338,6 +338,23 @@ public:
     return *this;
   }
 
+  /// test - Check if (This - RHS) is zero.
+  /// This is the same as reset(RHS) and any().
+  bool test(const BitVector &RHS) const {
+    unsigned ThisWords = NumBitWords(size());
+    unsigned RHSWords  = NumBitWords(RHS.size());
+    unsigned i;
+    for (i = 0; i != std::min(ThisWords, RHSWords); ++i)
+      if ((Bits[i] & ~RHS.Bits[i]) != 0)
+        return true;
+
+    for (; i != ThisWords ; ++i)
+      if (Bits[i] != 0)
+        return true;
+
+    return false;
+  }
+
   BitVector &operator|=(const BitVector &RHS) {
     if (size() < RHS.size())
       resize(RHS.size());

include/llvm/CodeGen/ISDOpcodes.h

@@ -637,6 +637,10 @@ namespace ISD {
     ATOMIC_LOAD_UMIN,
     ATOMIC_LOAD_UMAX,
 
+    /// This corresponds to the llvm.lifetime.* intrinsics. The first operand
+    /// is the chain and the second operand is the alloca pointer.
+    LIFETIME_START, LIFETIME_END,
+
     /// BUILTIN_OP_END - This must be the last enum value in this list.
     /// The target-specific pre-isel opcode values start here.
     BUILTIN_OP_END

include/llvm/CodeGen/MachineFrameInfo.h

@@ -28,6 +28,7 @@ class MachineFunction;
 class MachineBasicBlock;
 class TargetFrameLowering;
 class BitVector;
+class Value;
 
 /// The CalleeSavedInfo class tracks the information need to locate where a
 /// callee saved register is in the current frame.
@@ -103,14 +104,18 @@ class MachineFrameInfo {
     // protector.
     bool MayNeedSP;
 
+    /// Alloca - If this stack object is originated from an Alloca instruction
+    /// this value saves the original IR allocation. Can be NULL.
+    const Value *Alloca;
+
     // PreAllocated - If true, the object was mapped into the local frame
     // block and doesn't need additional handling for allocation beyond that.
     bool PreAllocated;
 
     StackObject(uint64_t Sz, unsigned Al, int64_t SP, bool IM,
-                bool isSS, bool NSP)
+                bool isSS, bool NSP, const Value *Val)
       : SPOffset(SP), Size(Sz), Alignment(Al), isImmutable(IM),
-        isSpillSlot(isSS), MayNeedSP(NSP), PreAllocated(false) {}
+        isSpillSlot(isSS), MayNeedSP(NSP), Alloca(Val), PreAllocated(false) {}
   };
 
   /// Objects - The list of stack objects allocated...
@@ -362,6 +367,14 @@ public:
     ensureMaxAlignment(Align);
   }
 
+  /// getObjectAllocation - Return the underlying Alloca of the specified
+  /// stack object if it exists. Returns 0 if none exists.
+  const Value* getObjectAllocation(int ObjectIdx) const {
+    assert(unsigned(ObjectIdx+NumFixedObjects) < Objects.size() &&
+           "Invalid Object Idx!");
+    return Objects[ObjectIdx+NumFixedObjects].Alloca;
+  }
+
   /// NeedsStackProtector - Returns true if the object may need stack
   /// protectors.
   bool MayNeedStackProtector(int ObjectIdx) const {
@@ -482,9 +495,10 @@ public:
   /// a nonnegative identifier to represent it.
   ///
   int CreateStackObject(uint64_t Size, unsigned Alignment, bool isSS,
-                        bool MayNeedSP = false) {
+                        bool MayNeedSP = false, const Value *Alloca = 0) {
     assert(Size != 0 && "Cannot allocate zero size stack objects!");
-    Objects.push_back(StackObject(Size, Alignment, 0, false, isSS, MayNeedSP));
+    Objects.push_back(StackObject(Size, Alignment, 0, false, isSS, MayNeedSP,
+                                  Alloca));
     int Index = (int)Objects.size() - NumFixedObjects - 1;
     assert(Index >= 0 && "Bad frame index!");
     ensureMaxAlignment(Alignment);
@@ -516,7 +530,7 @@ public:
   ///
   int CreateVariableSizedObject(unsigned Alignment) {
     HasVarSizedObjects = true;
-    Objects.push_back(StackObject(0, Alignment, 0, false, false, true));
+    Objects.push_back(StackObject(0, Alignment, 0, false, false, true, 0));
     ensureMaxAlignment(Alignment);
     return (int)Objects.size()-NumFixedObjects-1;
   }

include/llvm/CodeGen/Passes.h

@@ -404,6 +404,10 @@ namespace llvm {
   /// inserting cmov instructions.
   extern char &EarlyIfConverterID;
 
+  /// StackSlotColoring - This pass performs stack coloring and merging.
+  /// It merges disjoint allocas to reduce the stack size.
+  extern char &StackColoringID;
+
   /// IfConverter - This pass performs machine code if conversion.
   extern char &IfConverterID;
 

include/llvm/InitializePasses.h

@@ -232,6 +232,7 @@ void initializeSinkingPass(PassRegistry&);
 void initializeSlotIndexesPass(PassRegistry&);
 void initializeSpillPlacementPass(PassRegistry&);
 void initializeStackProtectorPass(PassRegistry&);
+void initializeStackColoringPass(PassRegistry&);
 void initializeStackSlotColoringPass(PassRegistry&);
 void initializeStripDeadDebugInfoPass(PassRegistry&);
 void initializeStripDeadPrototypesPassPass(PassRegistry&);

include/llvm/Target/Target.td

@@ -745,6 +745,18 @@ def BUNDLE : Instruction {
   let InOperandList = (ins variable_ops);
   let AsmString = "BUNDLE";
 }
+def LIFETIME_START : Instruction {
+  let OutOperandList = (outs);
+  let InOperandList = (ins i32imm:$id);
+  let AsmString = "LIFETIME_START";
+  let neverHasSideEffects = 1;
+}
+def LIFETIME_END : Instruction {
+  let OutOperandList = (outs);
+  let InOperandList = (ins i32imm:$id);
+  let AsmString = "LIFETIME_END";
+  let neverHasSideEffects = 1;
+}
 }
 
 //===----------------------------------------------------------------------===//

include/llvm/Target/TargetOpcodes.h

@@ -87,7 +87,11 @@ namespace TargetOpcode {
     /// BUNDLE - This instruction represents an instruction bundle. Instructions
     /// which immediately follow a BUNDLE instruction which are marked with
     /// 'InsideBundle' flag are inside the bundle.
-    BUNDLE
+    BUNDLE = 14,
+
+    /// Lifetime markers.
+    LIFETIME_START = 15,
+    LIFETIME_END = 16
   };
 } // end namespace TargetOpcode
 } // end namespace llvm

lib/CodeGen/CMakeLists.txt

@@ -95,6 +95,7 @@ add_llvm_library(LLVMCodeGen
   SplitKit.cpp
   StackProtector.cpp
   StackSlotColoring.cpp
+  StackColoring.cpp
   StrongPHIElimination.cpp
   TailDuplication.cpp
   TargetFrameLoweringImpl.cpp

lib/CodeGen/CodeGen.cpp

@@ -56,6 +56,7 @@ void llvm::initializeCodeGen(PassRegistry &Registry) {
   initializeRegisterCoalescerPass(Registry);
   initializeSlotIndexesPass(Registry);
   initializeStackProtectorPass(Registry);
+  initializeStackColoringPass(Registry);
   initializeStackSlotColoringPass(Registry);
   initializeStrongPHIEliminationPass(Registry);
   initializeTailDuplicatePassPass(Registry);

lib/CodeGen/MachineFunction.cpp

@@ -458,7 +458,9 @@ int MachineFrameInfo::CreateFixedObject(uint64_t Size, int64_t SPOffset,
   unsigned StackAlign = TFI.getStackAlignment();
   unsigned Align = MinAlign(SPOffset, StackAlign);
   Objects.insert(Objects.begin(), StackObject(Size, Align, SPOffset, Immutable,
-                                              /*isSS*/false, false));
+                                              /*isSS*/   false,
+                                              /*NeedSP*/ false,
+                                              /*Alloca*/ 0));
   return -++NumFixedObjects;
 }
 

lib/CodeGen/Passes.cpp

@@ -529,6 +529,10 @@ void TargetPassConfig::addMachineSSAOptimization() {
   // instructions dead.
   addPass(&OptimizePHIsID);
 
+  // This pass merges large allocas. StackSlotColoring is a different pass
+  // which merges spill slots.
+  addPass(&StackColoringID);
+
   // If the target requests it, assign local variables to stack slots relative
   // to one another and simplify frame index references where possible.
   addPass(&LocalStackSlotAllocationID);

lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp

@@ -97,7 +97,7 @@ void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf) {
             cast<ArrayType>(Ty)->getElementType()->isIntegerTy(8)));
         StaticAllocaMap[AI] =
           MF->getFrameInfo()->CreateStackObject(TySize, Align, false,
-                                                MayNeedSP);
+                                                MayNeedSP, AI);
       }
 
   for (; BB != EB; ++BB)

lib/CodeGen/SelectionDAG/InstrEmitter.cpp

@@ -873,6 +873,17 @@ EmitSpecialNode(SDNode *Node, bool IsClone, bool IsCloned,
     break;
   }
 
+  case ISD::LIFETIME_START:
+  case ISD::LIFETIME_END: {
+    unsigned TarOp = (Node->getOpcode() == ISD::LIFETIME_START) ?
+    TargetOpcode::LIFETIME_START : TargetOpcode::LIFETIME_END;
+
+    FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Node->getOperand(1));
+    BuildMI(*MBB, InsertPos, Node->getDebugLoc(), TII->get(TarOp))
+    .addFrameIndex(FI->getIndex());
+    break;
+  }
+
   case ISD::INLINEASM: {
     unsigned NumOps = Node->getNumOperands();
     if (Node->getOperand(NumOps-1).getValueType() == MVT::Glue)

lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp

@@ -656,6 +656,8 @@ void ScheduleDAGRRList::EmitNode(SUnit *SU) {
     break;
   case ISD::MERGE_VALUES:
   case ISD::TokenFactor:
+  case ISD::LIFETIME_START:
+  case ISD::LIFETIME_END:
   case ISD::CopyToReg:
   case ISD::CopyFromReg:
   case ISD::EH_LABEL:

lib/CodeGen/SelectionDAG/SelectionDAG.cpp

@@ -4161,6 +4161,8 @@ SelectionDAG::getMemIntrinsicNode(unsigned Opcode, DebugLoc dl, SDVTList VTList,
   assert((Opcode == ISD::INTRINSIC_VOID ||
           Opcode == ISD::INTRINSIC_W_CHAIN ||
           Opcode == ISD::PREFETCH ||
+          Opcode == ISD::LIFETIME_START ||
+          Opcode == ISD::LIFETIME_END ||
           (Opcode <= INT_MAX &&
            (int)Opcode >= ISD::FIRST_TARGET_MEMORY_OPCODE)) &&
          "Opcode is not a memory-accessing opcode!");

lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

@@ -19,6 +19,7 @@
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/Analysis/AliasAnalysis.h"
 #include "llvm/Analysis/ConstantFolding.h"
+#include "llvm/Analysis/ValueTracking.h"
 #include "llvm/Constants.h"
 #include "llvm/CallingConv.h"
 #include "llvm/DebugInfo.h"
@@ -5215,14 +5216,30 @@ SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, unsigned Intrinsic) {
                                         rw==1)); /* write */
     return 0;
   }
-
-  case Intrinsic::invariant_start:
   case Intrinsic::lifetime_start:
+  case Intrinsic::lifetime_end: {
+    SDValue Ops[2];
+    AllocaInst *LifetimeObject =dyn_cast_or_null<AllocaInst>(
+                                   GetUnderlyingObject(I.getArgOperand(1), TD));
+    // Could not find an Alloca.
+    if (!LifetimeObject)
+      return 0;
+
+    int FI = FuncInfo.StaticAllocaMap[LifetimeObject];
+    Ops[0] = getRoot();
+    Ops[1] = DAG.getFrameIndex(FI, TLI.getPointerTy(), true);
+    bool IsStart = (Intrinsic == Intrinsic::lifetime_start);
+    unsigned Opcode = (IsStart ? ISD::LIFETIME_START : ISD::LIFETIME_END);
+
+    Res = DAG.getNode(Opcode, dl, MVT::Other, Ops, 2);
+    DAG.setRoot(Res);
+    return 0;
+  }
+  case Intrinsic::invariant_start:
     // Discard region information.
     setValue(&I, DAG.getUNDEF(TLI.getPointerTy()));
     return 0;
   case Intrinsic::invariant_end:
-  case Intrinsic::lifetime_end:
     // Discard region information.
     return 0;
   case Intrinsic::donothing:

lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp

@@ -267,6 +267,8 @@ std::string SDNode::getOperationName(const SelectionDAG *G) const {
   case ISD::STACKRESTORE:               return "stackrestore";
   case ISD::TRAP:                       return "trap";
   case ISD::DEBUGTRAP:                  return "debugtrap";
+  case ISD::LIFETIME_START:             return "lifetime.start";
+  case ISD::LIFETIME_END:               return "lifetime.end";
 
   // Bit manipulation
   case ISD::BSWAP:                      return "bswap";

lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp

@@ -1801,10 +1801,13 @@ WalkChainUsers(const SDNode *ChainedNode,
         User->getOpcode() == ISD::HANDLENODE)  // Root of the graph.
       continue;
 
-    if (User->getOpcode() == ISD::CopyToReg ||
-        User->getOpcode() == ISD::CopyFromReg ||
-        User->getOpcode() == ISD::INLINEASM ||
-        User->getOpcode() == ISD::EH_LABEL) {
+    unsigned UserOpcode = User->getOpcode();
+    if (UserOpcode == ISD::CopyToReg ||
+        UserOpcode == ISD::CopyFromReg ||
+        UserOpcode == ISD::INLINEASM ||
+        UserOpcode == ISD::EH_LABEL ||
+        UserOpcode == ISD::LIFETIME_START ||
+        UserOpcode == ISD::LIFETIME_END) {
       // If their node ID got reset to -1 then they've already been selected.
       // Treat them like a MachineOpcode.
       if (User->getNodeId() == -1)
@@ -2220,6 +2223,8 @@ SelectCodeCommon(SDNode *NodeToMatch, const unsigned char *MatcherTable,
   case ISD::CopyFromReg:
   case ISD::CopyToReg:
   case ISD::EH_LABEL:
+  case ISD::LIFETIME_START:
+  case ISD::LIFETIME_END:
     NodeToMatch->setNodeId(-1); // Mark selected.
     return 0;
   case ISD::AssertSext:

lib/CodeGen/StackColoring.cpp

@@ -0,0 +1,657 @@
+//===-- StackColoring.cpp -------------------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This pass implements the stack-coloring optimization that looks for
+// lifetime markers machine instructions (LIFESTART_BEGIN and LIFESTART_END),
+// which represent the possible lifetime of stack slots. It attempts to
+// merge disjoint stack slots and reduce the used stack space.
+// NOTE: This pass is not StackSlotColoring, which optimizes spill slots.
+//
+// TODO: In the future we plan to improve stack coloring in the following ways:
+// 1. Allow merging multiple small slots into a single larger slot at different
+//    offsets.
+// 2. Merge this pass with StackSlotColoring and allow merging of allocas with
+//    spill slots.
+//
+//===----------------------------------------------------------------------===//
+
+#define DEBUG_TYPE "stackcoloring"
+#include "MachineTraceMetrics.h"
+#include "llvm/Function.h"
+#include "llvm/Module.h"
+#include "llvm/ADT/BitVector.h"
+#include "llvm/Analysis/Dominators.h"
+#include "llvm/Analysis/ValueTracking.h"
+#include "llvm/ADT/DepthFirstIterator.h"
+#include "llvm/ADT/PostOrderIterator.h"
+#include "llvm/ADT/SetVector.h"
+#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/SparseSet.h"
+#include "llvm/ADT/Statistic.h"
+#include "llvm/CodeGen/LiveInterval.h"
+#include "llvm/CodeGen/MachineLoopInfo.h"
+#include "llvm/CodeGen/MachineBranchProbabilityInfo.h"
+#include "llvm/CodeGen/MachineDominators.h"
+#include "llvm/CodeGen/MachineBasicBlock.h"
+#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/MachineLoopInfo.h"
+#include "llvm/CodeGen/MachineModuleInfo.h"
+#include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/MachineFrameInfo.h"
+#include "llvm/CodeGen/MachineMemOperand.h"
+#include "llvm/CodeGen/Passes.h"
+#include "llvm/CodeGen/SlotIndexes.h"
+#include "llvm/DebugInfo.h"
+#include "llvm/MC/MCInstrItineraries.h"
+#include "llvm/Target/TargetInstrInfo.h"
+#include "llvm/Target/TargetRegisterInfo.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/raw_ostream.h"
+
+using namespace llvm;
+
+static cl::opt<bool>
+DisableColoring("no-stack-coloring",
+               cl::init(false), cl::Hidden,
+               cl::desc("Suppress stack coloring"));
+
+STATISTIC(NumMarkerSeen,  "Number of life markers found.");
+STATISTIC(StackSpaceSaved, "Number of bytes saved due to merging slots.");
+STATISTIC(StackSlotMerged, "Number of stack slot merged.");
+
+//===----------------------------------------------------------------------===//
+//                           StackColoring Pass
+//===----------------------------------------------------------------------===//
+
+namespace {
+/// StackColoring - A machine pass for merging disjoint stack allocations,
+/// marked by the LIFETIME_START and LIFETIME_END pseudo instructions.
+class StackColoring : public MachineFunctionPass {
+  MachineFrameInfo *MFI;
+  MachineFunction *MF;
+
+  /// A class representing liveness information for a single basic block.
+  /// Each bit in the BitVector represents the liveness property
+  /// for a different stack slot.
+  struct BlockLifetimeInfo {
+    /// Which slots BEGINs in each basic block.
+    BitVector Begin;
+    /// Which slots ENDs in each basic block.
+    BitVector End;
+    /// Which slots are marked as LIVE_IN, coming into each basic block.
+    BitVector LiveIn;
+    /// Which slots are marked as LIVE_OUT, coming out of each basic block.
+    BitVector LiveOut;
+  };
+
+  /// Maps active slots (per bit) for each basic block.
+  DenseMap<MachineBasicBlock*, BlockLifetimeInfo> BlockLiveness;
+
+  /// Maps serial numbers to basic blocks.
+  DenseMap<MachineBasicBlock*, int> BasicBlocks;
+  /// Maps basic blocks to a serial number.
+  SmallVector<MachineBasicBlock*, 8> BasicBlockNumbering;
+
+  /// Maps liveness intervals for each slot.
+  SmallVector<LiveInterval*, 16> Intervals;
+  /// VNInfo is used for the construction of LiveIntervals.
+  VNInfo::Allocator VNInfoAllocator;
+  /// SlotIndex analysis object.
+  SlotIndexes* Indexes;
+
+  /// The list of lifetime markers found. These markers are to be removed
+  /// once the coloring is done.
+  SmallVector<MachineInstr*, 8> Markers;
+
+  /// SlotSizeSorter - A Sort utility for arranging stack slots according
+  /// to their size.
+  struct SlotSizeSorter {
+    MachineFrameInfo *MFI;
+    SlotSizeSorter(MachineFrameInfo *mfi) : MFI(mfi) { }
+    bool operator()(int LHS, int RHS) {
+      // We use -1 to denote a uninteresting slot. Place these slots at the end.
+      if (LHS == -1) return false;
+      if (RHS == -1) return true;
+      // Sort according to size.
+      return MFI->getObjectSize(LHS) > MFI->getObjectSize(RHS);
+  }
+};
+
+public:
+  static char ID;
+  StackColoring() : MachineFunctionPass(ID) {
+    initializeStackColoringPass(*PassRegistry::getPassRegistry());
+  }
+  void getAnalysisUsage(AnalysisUsage &AU) const;
+  bool runOnMachineFunction(MachineFunction &MF);
+
+private:
+  /// Debug.
+  void dump();
+
+  /// Removes all of the lifetime marker instructions from the function.
+  /// \returns true if any markers were removed.
+  bool removeAllMarkers();
+
+  /// Scan the machine function and find all of the lifetime markers.
+  /// Record the findings in the BEGIN and END vectors.
+  /// \returns the number of markers found.
+  unsigned collectMarkers(unsigned NumSlot);
+
+  /// Perform the dataflow calculation and calculate the lifetime for each of
+  /// the slots, based on the BEGIN/END vectors. Set the LifetimeLIVE_IN and
+  /// LifetimeLIVE_OUT maps that represent which stack slots are live coming
+  /// in and out blocks.
+  void calculateLocalLiveness();
+
+  /// Construct the LiveIntervals for the slots.
+  void calculateLiveIntervals(unsigned NumSlots);
+
+  /// Go over the machine function and change instructions which use stack
+  /// slots to use the joint slots.
+  void remapInstructions(DenseMap<int, int> &SlotRemap);
+
+  /// Map entries which point to other entries to their destination.
+  ///   A->B->C becomes A->C.
+   void expungeSlotMap(DenseMap<int, int> &SlotRemap, unsigned NumSlots);
+};
+} // end anonymous namespace
+
+char StackColoring::ID = 0;
+char &llvm::StackColoringID = StackColoring::ID;
+
+INITIALIZE_PASS_BEGIN(StackColoring,
+                   "stack-coloring", "Merge disjoint stack slots", false, false)
+INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree)
+INITIALIZE_PASS_DEPENDENCY(SlotIndexes)
+INITIALIZE_PASS_END(StackColoring,
+                   "stack-coloring", "Merge disjoint stack slots", false, false)
+
+void StackColoring::getAnalysisUsage(AnalysisUsage &AU) const {
+  AU.addRequired<MachineDominatorTree>();
+  AU.addPreserved<MachineDominatorTree>();
+  AU.addRequired<SlotIndexes>();
+  MachineFunctionPass::getAnalysisUsage(AU);
+}
+
+void StackColoring::dump() {
+  for (df_iterator<MachineFunction*> FI = df_begin(MF), FE = df_end(MF);
+       FI != FE; ++FI) {
+    unsigned Num = BasicBlocks[*FI];
+    DEBUG(dbgs()<<"Inspecting block #"<<Num<<" ["<<FI->getName()<<"]\n");
+    Num = 0;
+    DEBUG(dbgs()<<"BEGIN  : {");
+    for (unsigned i=0; i < BlockLiveness[*FI].Begin.size(); ++i)
+      DEBUG(dbgs()<<BlockLiveness[*FI].Begin.test(i)<<" ");
+    DEBUG(dbgs()<<"}\n");
+
+    DEBUG(dbgs()<<"END    : {");
+    for (unsigned i=0; i < BlockLiveness[*FI].End.size(); ++i)
+      DEBUG(dbgs()<<BlockLiveness[*FI].End.test(i)<<" ");
+
+    DEBUG(dbgs()<<"}\n");
+
+    DEBUG(dbgs()<<"LIVE_IN: {");
+    for (unsigned i=0; i < BlockLiveness[*FI].LiveIn.size(); ++i)
+      DEBUG(dbgs()<<BlockLiveness[*FI].LiveIn.test(i)<<" ");
+
+    DEBUG(dbgs()<<"}\n");
+    DEBUG(dbgs()<<"LIVEOUT: {");
+    for (unsigned i=0; i < BlockLiveness[*FI].LiveOut.size(); ++i)
+      DEBUG(dbgs()<<BlockLiveness[*FI].LiveOut.test(i)<<" ");
+    DEBUG(dbgs()<<"}\n");
+  }
+}
+
+unsigned StackColoring::collectMarkers(unsigned NumSlot) {
+  unsigned MarkersFound = 0;
+  // Scan the function to find all lifetime markers.
+  // NOTE: We use the a reverse-post-order iteration to ensure that we obtain a
+  // deterministic numbering, and because we'll need a post-order iteration
+  // later for solving the liveness dataflow problem.
+  for (df_iterator<MachineFunction*> FI = df_begin(MF), FE = df_end(MF);
+       FI != FE; ++FI) {
+
+    // Assign a serial number to this basic block.
+    BasicBlocks[*FI] = BasicBlockNumbering.size();;
+    BasicBlockNumbering.push_back(*FI);
+
+    BlockLiveness[*FI].Begin.resize(NumSlot);
+    BlockLiveness[*FI].End.resize(NumSlot);
+
+    for (MachineBasicBlock::iterator BI = (*FI)->begin(), BE = (*FI)->end();
+         BI != BE; ++BI) {
+
+      if (BI->getOpcode() != TargetOpcode::LIFETIME_START &&
+          BI->getOpcode() != TargetOpcode::LIFETIME_END)
+        continue;
+
+      Markers.push_back(BI);
+
+      bool IsStart = BI->getOpcode() == TargetOpcode::LIFETIME_START;
+      MachineOperand &MI = BI->getOperand(0);
+      unsigned Slot = MI.getIndex();
+
+      MarkersFound++;
+
+      const Value* Allocation = MFI->getObjectAllocation(Slot);
+      if (Allocation) {
+        DEBUG(dbgs()<<"Found lifetime marker for allocation: "<<
+              Allocation->getName()<<"\n");
+      }
+
+      if (IsStart) {
+        BlockLiveness[*FI].Begin.set(Slot);
+      } else {
+        if (BlockLiveness[*FI].Begin.test(Slot)) {
+          // Allocas that start and end within a single block are handled
+          // specially when computing the LiveIntervals to avoid pessimizing
+          // the liveness propagation.
+          BlockLiveness[*FI].Begin.reset(Slot);
+        } else {
+          BlockLiveness[*FI].End.set(Slot);
+        }
+      }
+    }
+  }
+
+  // Update statistics.
+  NumMarkerSeen += MarkersFound;
+  return MarkersFound;
+}
+
+void StackColoring::calculateLocalLiveness() {
+  // Perform a standard reverse dataflow computation to solve for
+  // global liveness.  The BEGIN set here is equivalent to KILL in the standard
+  // formulation, and END is equivalent to GEN.  The result of this computation
+  // is a map from blocks to bitvectors where the bitvectors represent which
+  // allocas are live in/out of that block.
+  SmallPtrSet<MachineBasicBlock*, 8> BBSet(BasicBlockNumbering.begin(),
+                                           BasicBlockNumbering.end());
+  unsigned NumSSMIters = 0;
+  bool changed = true;
+  while (changed) {
+    changed = false;
+    ++NumSSMIters;
+
+    SmallPtrSet<MachineBasicBlock*, 8> NextBBSet;
+
+    for (SmallVector<MachineBasicBlock*, 8>::iterator
+         PI = BasicBlockNumbering.begin(), PE = BasicBlockNumbering.end();
+         PI != PE; ++PI) {
+
+      MachineBasicBlock *BB = *PI;
+      if (!BBSet.count(BB)) continue;
+
+      BitVector LocalLiveIn;
+      BitVector LocalLiveOut;
+
+      // Forward propagation from begins to ends.
+      for (MachineBasicBlock::pred_iterator PI = BB->pred_begin(),
+           PE = BB->pred_end(); PI != PE; ++PI)
+        LocalLiveIn |= BlockLiveness[*PI].LiveOut;
+      LocalLiveIn |= BlockLiveness[BB].End;
+      LocalLiveIn.reset(BlockLiveness[BB].Begin);
+
+      // Reverse propagation from ends to begins.
+      for (MachineBasicBlock::succ_iterator SI = BB->succ_begin(),
+           SE = BB->succ_end(); SI != SE; ++SI)
+        LocalLiveOut |= BlockLiveness[*SI].LiveIn;
+      LocalLiveOut |= BlockLiveness[BB].Begin;
+      LocalLiveOut.reset(BlockLiveness[BB].End);
+
+      LocalLiveIn |= LocalLiveOut;
+      LocalLiveOut |= LocalLiveIn;
+
+      // After adopting the live bits, we need to turn-off the bits which
+      // are de-activated in this block.
+      LocalLiveOut.reset(BlockLiveness[BB].End);
+      LocalLiveIn.reset(BlockLiveness[BB].Begin);
+
+      if (LocalLiveIn.test(BlockLiveness[BB].LiveIn)) {
+        changed = true;
+        BlockLiveness[BB].LiveIn |= LocalLiveIn;
+
+        for (MachineBasicBlock::pred_iterator PI = BB->pred_begin(),
+             PE = BB->pred_end(); PI != PE; ++PI)
+          NextBBSet.insert(*PI);
+      }
+
+      if (LocalLiveOut.test(BlockLiveness[BB].LiveOut)) {
+        changed = true;
+        BlockLiveness[BB].LiveOut |= LocalLiveOut;
+
+        for (MachineBasicBlock::succ_iterator SI = BB->succ_begin(),
+             SE = BB->succ_end(); SI != SE; ++SI)
+          NextBBSet.insert(*SI);
+      }
+    }
+
+    BBSet = NextBBSet;
+  }// while changed.
+}
+
+void StackColoring::calculateLiveIntervals(unsigned NumSlots) {
+  SmallVector<SlotIndex, 16> Starts;
+  SmallVector<SlotIndex, 16> Finishes;
+
+  // For each block, find which slots are active within this block
+  // and update the live intervals.
+  for (MachineFunction::iterator MBB = MF->begin(), MBBe = MF->end();
+       MBB != MBBe; ++MBB) {
+    Starts.clear();
+    Starts.resize(NumSlots);
+    Finishes.clear();
+    Finishes.resize(NumSlots);
+
+    BitVector Alive = BlockLiveness[MBB].LiveIn;
+    Alive |= BlockLiveness[MBB].LiveOut;
+
+    if (Alive.any()) {
+      for (int pos = Alive.find_first(); pos != -1;
+           pos = Alive.find_next(pos)) {
+        Starts[pos] = Indexes->getMBBStartIdx(MBB);
+        Finishes[pos] = Indexes->getMBBEndIdx(MBB);
+      }
+    }
+
+    for (SmallVector<MachineInstr*, 8>::iterator it = Markers.begin(),
+         e = Markers.end(); it != e; ++it) {
+      MachineInstr *MI = *it;
+      assert((MI->getOpcode() == TargetOpcode::LIFETIME_START ||
+              MI->getOpcode() == TargetOpcode::LIFETIME_END) &&
+             "Invalid Lifetime marker");
+
+      if (MI->getParent() == MBB) {
+        bool IsStart = MI->getOpcode() == TargetOpcode::LIFETIME_START;
+        MachineOperand &Mo = MI->getOperand(0);
+        int Slot = Mo.getIndex();
+        assert(Slot >= 0 && "Invalid slot");
+        if (IsStart) {
+          Starts[Slot] = Indexes->getInstructionIndex(MI);
+        } else {
+          Finishes[Slot] = Indexes->getInstructionIndex(MI);
+        }
+      }
+    }
+
+    for (unsigned i = 0; i < NumSlots; ++i) {
+      assert(!!Starts[i] == !!Finishes[i] && "Unmatched range");
+      if (Starts[i] == Finishes[i])
+        continue;
+
+      assert(Starts[i] && Finishes[i] && "Invalid interval");
+      VNInfo *ValNum = Intervals[i]->getValNumInfo(0);
+      SlotIndex S = Starts[i];
+      SlotIndex F = Finishes[i];
+      if (S < F) {
+        // We have a single consecutive region.
+        Intervals[i]->addRange(LiveRange(S, F, ValNum));
+      } else {
+        // We have two non consecutive regions. This happens when
+        // LIFETIME_START appears after the LIFETIME_END marker.
+        SlotIndex NewStart = Indexes->getMBBStartIdx(MBB);
+        SlotIndex NewFin = Indexes->getMBBEndIdx(MBB);
+        Intervals[i]->addRange(LiveRange(NewStart, F, ValNum));
+        Intervals[i]->addRange(LiveRange(S, NewFin, ValNum));
+      }
+    }
+  }
+}
+
+bool StackColoring::removeAllMarkers() {
+  unsigned Count = 0;
+  for (unsigned i = 0; i < Markers.size(); ++i) {
+    Markers[i]->eraseFromParent();
+    Count++;
+  }
+  Markers.clear();
+
+  DEBUG(dbgs()<<"Removed "<<Count<<" markers.\n");
+  return Count;
+}
+
+void StackColoring::remapInstructions(DenseMap<int, int> &SlotRemap) {
+  unsigned FixedInstr = 0;
+  unsigned FixedMemOp = 0;
+  unsigned FixedDbg = 0;
+  MachineModuleInfo *MMI = &MF->getMMI();
+
+  // Remap debug information that refers to stack slots.
+  MachineModuleInfo::VariableDbgInfoMapTy &VMap = MMI->getVariableDbgInfo();
+  for (MachineModuleInfo::VariableDbgInfoMapTy::iterator VI = VMap.begin(),
+       VE = VMap.end(); VI != VE; ++VI) {
+    const MDNode *Var = VI->first;
+    if (!Var) continue;
+    std::pair<unsigned, DebugLoc> &VP = VI->second;
+    if (SlotRemap.count(VP.first)) {
+      DEBUG(dbgs()<<"Remapping debug info for ["<<Var->getName()<<"].\n");
+      VP.first = SlotRemap[VP.first];
+      FixedDbg++;
+    }
+  }
+
+  // Keep a list of *allocas* which need to be remapped.
+  DenseMap<const Value*, const Value*> Allocas;
+  for (DenseMap<int, int>::iterator it = SlotRemap.begin(),
+       e = SlotRemap.end(); it != e; ++it) {
+    const Value* From = MFI->getObjectAllocation(it->first);
+    const Value* To = MFI->getObjectAllocation(it->second);
+    assert(To && From && "Invalid allocation object");
+    Allocas[From] = To;
+  }
+
+  // Remap all instructions to the new stack slots.
+  MachineFunction::iterator BB, BBE;
+  MachineBasicBlock::iterator I, IE;
+  for (BB = MF->begin(), BBE = MF->end(); BB != BBE; ++BB)
+    for (I = BB->begin(), IE = BB->end(); I != IE; ++I) {
+
+      // Update the MachineMemOperand to use the new alloca.
+      for (MachineInstr::mmo_iterator MM = I->memoperands_begin(),
+           E = I->memoperands_end(); MM != E; ++MM) {
+        MachineMemOperand *MMO = *MM;
+
+        const Value *V = MMO->getValue();
+
+        if (!V)
+          continue;
+
+        // Climb up and find the original alloca.
+        V = GetUnderlyingObject(V);
+        // If we did not find one, or if the one that we found is not in our
+        // map, then move on.
+        if (!V || !Allocas.count(V))
+          continue;
+
+        MMO->setValue(Allocas[V]);
+        FixedMemOp++;
+      }
+
+      // Update all of the machine instruction operands.
+      for (unsigned i = 0 ; i <  I->getNumOperands(); ++i) {
+        MachineOperand &MO = I->getOperand(i);
+
+        if (!MO.isFI())
+          continue;
+        int FromSlot = MO.getIndex();
+
+        // Don't touch arguments.
+        if (FromSlot<0)
+          continue;
+
+        // Only look at mapped slots.
+        if (!SlotRemap.count(FromSlot))
+          continue;
+
+        // Fix the machine instructions.
+        int ToSlot = SlotRemap[FromSlot];
+        MO.setIndex(ToSlot);
+        FixedInstr++;
+      }
+    }
+
+  DEBUG(dbgs()<<"Fixed "<<FixedMemOp<<" machine memory operands.\n");
+  DEBUG(dbgs()<<"Fixed "<<FixedDbg<<" debug locations.\n");
+  DEBUG(dbgs()<<"Fixed "<<FixedInstr<<" machine instructions.\n");
+}
+
+void StackColoring::expungeSlotMap(DenseMap<int, int> &SlotRemap,
+                                   unsigned NumSlots) {
+  // Expunge slot remap map.
+  for (unsigned i=0; i < NumSlots; ++i) {
+    // If we are remapping i
+    if (SlotRemap.count(i)) {
+      int Target = SlotRemap[i];
+      // As long as our target is mapped to something else, follow it.
+      while (SlotRemap.count(Target)) {
+        Target = SlotRemap[Target];
+        SlotRemap[i] = Target;
+      }
+    }
+  }
+}
+
+bool StackColoring::runOnMachineFunction(MachineFunction &Func) {
+  DEBUG(dbgs() << "********** Stack Coloring **********\n"
+               << "********** Function: "
+               << ((Value*)Func.getFunction())->getName() << '\n');
+  MF = &Func;
+  MFI = MF->getFrameInfo();
+  Indexes = &getAnalysis<SlotIndexes>();
+  BlockLiveness.clear();
+  BasicBlocks.clear();
+  BasicBlockNumbering.clear();
+  Markers.clear();
+  Intervals.clear();
+  VNInfoAllocator.Reset();
+
+  unsigned NumSlots = MFI->getObjectIndexEnd();
+
+  // If there are no stack slots then there are no markers to remove.
+  if (!NumSlots)
+    return false;
+
+  SmallVector<int, 8> SortedSlots;
+
+  SortedSlots.reserve(NumSlots);
+  Intervals.reserve(NumSlots);
+
+  unsigned NumMarkers = collectMarkers(NumSlots);
+
+  unsigned TotalSize = 0;
+  DEBUG(dbgs()<<"Found "<<NumMarkers<<" markers and "<<NumSlots<<" slots\n");
+  DEBUG(dbgs()<<"Slot structure:\n");
+
+  for (int i=0; i < MFI->getObjectIndexEnd(); ++i) {
+    DEBUG(dbgs()<<"Slot #"<<i<<" - "<<MFI->getObjectSize(i)<<" bytes.\n");
+    TotalSize += MFI->getObjectSize(i);
+  }
+
+  DEBUG(dbgs()<<"Total Stack size: "<<TotalSize<<" bytes\n\n");
+
+  // Don't continue because there are not enough lifetime markers, or the
+  // stack or too small, or we are told not to optimize the slots.
+  if (NumMarkers < 2 || TotalSize < 16 || DisableColoring) {
+    DEBUG(dbgs()<<"Will not try to merge slots.\n");
+    return removeAllMarkers();
+  }
+
+  for (unsigned i=0; i < NumSlots; ++i) {
+    LiveInterval *LI = new LiveInterval(i, 0);
+    Intervals.push_back(LI);
+    LI->getNextValue(Indexes->getZeroIndex(), VNInfoAllocator);
+    SortedSlots.push_back(i);
+  }
+
+  // Calculate the liveness of each block.
+  calculateLocalLiveness();
+
+  // Propagate the liveness information.
+  calculateLiveIntervals(NumSlots);
+
+  // Maps old slots to new slots.
+  DenseMap<int, int> SlotRemap;
+  unsigned RemovedSlots = 0;
+  unsigned ReducedSize = 0;
+
+  // Do not bother looking at empty intervals.
+  for (unsigned I = 0; I < NumSlots; ++I) {
+    if (Intervals[SortedSlots[I]]->empty())
+      SortedSlots[I] = -1;
+  }
+
+  // This is a simple greedy algorithm for merging allocas. First, sort the
+  // slots, placing the largest slots first. Next, perform an n^2 scan and look
+  // for disjoint slots. When you find disjoint slots, merge the samller one
+  // into the bigger one and update the live interval. Remove the small alloca
+  // and continue.
+
+  // Sort the slots according to their size. Place unused slots at the end.
+  std::sort(SortedSlots.begin(), SortedSlots.end(), SlotSizeSorter(MFI));
+
+  bool Chanded = true;
+  while (Chanded) {
+    Chanded = false;
+    for (unsigned I = 0; I < NumSlots; ++I) {
+      if (SortedSlots[I] == -1)
+        continue;
+
+      for (unsigned J=0; J < NumSlots; ++J) {
+        if (SortedSlots[J] == -1)
+          continue;
+
+        int FirstSlot = SortedSlots[I];
+        int SecondSlot = SortedSlots[J];
+        LiveInterval *First = Intervals[FirstSlot];
+        LiveInterval *Second = Intervals[SecondSlot];
+        assert (!First->empty() && !Second->empty() && "Found an empty range");
+
+        // Merge disjoint slots.
+        if (!First->overlaps(*Second)) {
+          Chanded = true;
+          First->MergeRangesInAsValue(*Second, First->getValNumInfo(0));
+          SlotRemap[SecondSlot] = FirstSlot;
+          SortedSlots[J] = -1;
+          DEBUG(dbgs()<<"Merging #"<<I<<" and slots #"<<J<<" together.\n");
+          unsigned MaxAlignment = std::max(MFI->getObjectAlignment(FirstSlot),
+                                           MFI->getObjectAlignment(SecondSlot));
+
+          assert(MFI->getObjectSize(FirstSlot) >=
+                 MFI->getObjectSize(SecondSlot) &&
+                 "Merging a small object into a larger one");
+
+          RemovedSlots+=1;
+          ReducedSize += MFI->getObjectSize(SecondSlot);
+          MFI->setObjectAlignment(FirstSlot, MaxAlignment);
+          MFI->RemoveStackObject(SecondSlot);
+        }
+      }
+    }
+  }// While changed.
+
+  // Record statistics.
+  StackSpaceSaved += ReducedSize;
+  StackSlotMerged += RemovedSlots;
+  DEBUG(dbgs()<<"Merge "<<RemovedSlots<<" slots. Saved "<<
+        ReducedSize<<" bytes\n");
+
+  // Scan the entire function and update all machine operands that use frame
+  // indices to use the remapped frame index.
+  expungeSlotMap(SlotRemap, NumSlots);
+  remapInstructions(SlotRemap);
+
+  // Release the intervals.
+  for (unsigned I = 0; I < NumSlots; ++I) {
+    delete Intervals[I];
+  }
+
+  return removeAllMarkers();
+}

test/CodeGen/X86/StackColoring.ll

@@ -0,0 +1,272 @@
+; RUN: llc < %s                    | FileCheck %s --check-prefix=YESCOLOR
+; RUN: llc -no-stack-coloring < %s | FileCheck %s --check-prefix=NOCOLOR
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+;YESCOLOR: subq  $136, %rsp
+;NOCOLOR: subq  $264, %rsp
+
+
+define i32 @myCall_w2(i32 %in) {
+entry:
+  %a = alloca [17 x i8*], align 8
+  %a2 = alloca [16 x i8*], align 8
+  %b = bitcast [17 x i8*]* %a to i8*
+  %b2 = bitcast [16 x i8*]* %a2 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b)
+  %t1 = call i32 @foo(i32 %in, i8* %b)
+  %t2 = call i32 @foo(i32 %in, i8* %b)
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t3 = call i32 @foo(i32 %in, i8* %b2)
+  %t4 = call i32 @foo(i32 %in, i8* %b2)
+  call void @llvm.lifetime.end(i64 -1, i8* %b2)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+}
+
+
+;YESCOLOR: subq  $272, %rsp
+;NOCOLOR: subq  $272, %rsp
+
+define i32 @myCall2_no_merge(i32 %in, i1 %d) {
+entry:
+  %a = alloca [17 x i8*], align 8
+  %a2 = alloca [16 x i8*], align 8
+  %b = bitcast [17 x i8*]* %a to i8*
+  %b2 = bitcast [16 x i8*]* %a2 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b)
+  %t1 = call i32 @foo(i32 %in, i8* %b)
+  %t2 = call i32 @foo(i32 %in, i8* %b)
+  br i1 %d, label %bb2, label %bb3
+bb2:
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t3 = call i32 @foo(i32 %in, i8* %b2)
+  %t4 = call i32 @foo(i32 %in, i8* %b2)
+  call void @llvm.lifetime.end(i64 -1, i8* %b2)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  ret i32 %t7
+bb3:
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  ret i32 0
+}
+
+;YESCOLOR: subq  $144, %rsp
+;NOCOLOR: subq  $272, %rsp
+
+define i32 @myCall2_w2(i32 %in, i1 %d) {
+entry:
+  %a = alloca [17 x i8*], align 8
+  %a2 = alloca [16 x i8*], align 8
+  %b = bitcast [17 x i8*]* %a to i8*
+  %b2 = bitcast [16 x i8*]* %a2 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b)
+  %t1 = call i32 @foo(i32 %in, i8* %b)
+  %t2 = call i32 @foo(i32 %in, i8* %b)
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  br i1 %d, label %bb2, label %bb3
+bb2:
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t3 = call i32 @foo(i32 %in, i8* %b2)
+  %t4 = call i32 @foo(i32 %in, i8* %b2)
+  call void @llvm.lifetime.end(i64 -1, i8* %b2)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+bb3:
+  ret i32 0
+}
+;YESCOLOR: subq  $208, %rsp
+;NOCOLOR: subq  $400, %rsp
+
+
+
+
+define i32 @myCall_w4(i32 %in) {
+entry:
+  %a1 = alloca [14 x i8*], align 8
+  %a2 = alloca [13 x i8*], align 8
+  %a3 = alloca [12 x i8*], align 8
+  %a4 = alloca [11 x i8*], align 8
+  %b1 = bitcast [14 x i8*]* %a1 to i8*
+  %b2 = bitcast [13 x i8*]* %a2 to i8*
+  %b3 = bitcast [12 x i8*]* %a3 to i8*
+  %b4 = bitcast [11 x i8*]* %a4 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b4)
+  call void @llvm.lifetime.start(i64 -1, i8* %b1)
+  %t1 = call i32 @foo(i32 %in, i8* %b1)
+  %t2 = call i32 @foo(i32 %in, i8* %b1)
+  call void @llvm.lifetime.end(i64 -1, i8* %b1)
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t9 = call i32 @foo(i32 %in, i8* %b2)
+  %t8 = call i32 @foo(i32 %in, i8* %b2)
+  call void @llvm.lifetime.end(i64 -1, i8* %b2)
+  call void @llvm.lifetime.start(i64 -1, i8* %b3)
+  %t3 = call i32 @foo(i32 %in, i8* %b3)
+  %t4 = call i32 @foo(i32 %in, i8* %b3)
+  call void @llvm.lifetime.end(i64 -1, i8* %b3)
+  %t11 = call i32 @foo(i32 %in, i8* %b4)
+  call void @llvm.lifetime.end(i64 -1, i8* %b4)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+}
+
+;YESCOLOR: subq  $112, %rsp
+;NOCOLOR: subq  $400, %rsp
+
+define i32 @myCall2_w4(i32 %in) {
+entry:
+  %a1 = alloca [14 x i8*], align 8
+  %a2 = alloca [13 x i8*], align 8
+  %a3 = alloca [12 x i8*], align 8
+  %a4 = alloca [11 x i8*], align 8
+  %b1 = bitcast [14 x i8*]* %a1 to i8*
+  %b2 = bitcast [13 x i8*]* %a2 to i8*
+  %b3 = bitcast [12 x i8*]* %a3 to i8*
+  %b4 = bitcast [11 x i8*]* %a4 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b1)
+  %t1 = call i32 @foo(i32 %in, i8* %b1)
+  %t2 = call i32 @foo(i32 %in, i8* %b1)
+  call void @llvm.lifetime.end(i64 -1, i8* %b1)
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t9 = call i32 @foo(i32 %in, i8* %b2)
+  %t8 = call i32 @foo(i32 %in, i8* %b2)
+  call void @llvm.lifetime.end(i64 -1, i8* %b2)
+  call void @llvm.lifetime.start(i64 -1, i8* %b3)
+  %t3 = call i32 @foo(i32 %in, i8* %b3)
+  %t4 = call i32 @foo(i32 %in, i8* %b3)
+  call void @llvm.lifetime.end(i64 -1, i8* %b3)
+  br i1 undef, label %bb2, label %bb3
+bb2:
+  call void @llvm.lifetime.start(i64 -1, i8* %b4)
+  %t11 = call i32 @foo(i32 %in, i8* %b4)
+  call void @llvm.lifetime.end(i64 -1, i8* %b4)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+bb3:
+  ret i32 0
+}
+
+
+;YESCOLOR: subq  $144, %rsp
+;NOCOLOR: subq  $272, %rsp
+
+
+define i32 @myCall2_noend(i32 %in, i1 %d) {
+entry:
+  %a = alloca [17 x i8*], align 8
+  %a2 = alloca [16 x i8*], align 8
+  %b = bitcast [17 x i8*]* %a to i8*
+  %b2 = bitcast [16 x i8*]* %a2 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b)
+  %t1 = call i32 @foo(i32 %in, i8* %b)
+  %t2 = call i32 @foo(i32 %in, i8* %b)
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  br i1 %d, label %bb2, label %bb3
+bb2:
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t3 = call i32 @foo(i32 %in, i8* %b2)
+  %t4 = call i32 @foo(i32 %in, i8* %b2)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+bb3:
+  ret i32 0
+}
+
+;YESCOLOR: subq  $144, %rsp
+;NOCOLOR: subq  $272, %rsp
+define i32 @myCall2_noend2(i32 %in, i1 %d) {
+entry:
+  %a = alloca [17 x i8*], align 8
+  %a2 = alloca [16 x i8*], align 8
+  %b = bitcast [17 x i8*]* %a to i8*
+  %b2 = bitcast [16 x i8*]* %a2 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %b)
+  %t1 = call i32 @foo(i32 %in, i8* %b)
+  %t2 = call i32 @foo(i32 %in, i8* %b)
+  br i1 %d, label %bb2, label %bb3
+bb2:
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t3 = call i32 @foo(i32 %in, i8* %b2)
+  %t4 = call i32 @foo(i32 %in, i8* %b2)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+bb3:
+  ret i32 0
+}
+
+
+;YESCOLOR: subq  $144, %rsp
+;NOCOLOR: subq  $272, %rsp
+define i32 @myCall2_nostart(i32 %in, i1 %d) {
+entry:
+  %a = alloca [17 x i8*], align 8
+  %a2 = alloca [16 x i8*], align 8
+  %b = bitcast [17 x i8*]* %a to i8*
+  %b2 = bitcast [16 x i8*]* %a2 to i8*
+  %t1 = call i32 @foo(i32 %in, i8* %b)
+  %t2 = call i32 @foo(i32 %in, i8* %b)
+  call void @llvm.lifetime.end(i64 -1, i8* %b)
+  br i1 %d, label %bb2, label %bb3
+bb2:
+  call void @llvm.lifetime.start(i64 -1, i8* %b2)
+  %t3 = call i32 @foo(i32 %in, i8* %b2)
+  %t4 = call i32 @foo(i32 %in, i8* %b2)
+  %t5 = add i32 %t1, %t2
+  %t6 = add i32 %t3, %t4
+  %t7 = add i32 %t5, %t6
+  ret i32 %t7
+bb3:
+  ret i32 0
+}
+
+; Adopt the test from Transforms/Inline/array_merge.ll'
+;YESCOLOR: subq  $816, %rsp
+;NOCOLOR: subq  $1616, %rsp
+define void @array_merge() nounwind ssp {
+entry:
+  %A.i1 = alloca [100 x i32], align 4
+  %B.i2 = alloca [100 x i32], align 4
+  %A.i = alloca [100 x i32], align 4
+  %B.i = alloca [100 x i32], align 4
+  %0 = bitcast [100 x i32]* %A.i to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %0) nounwind
+  %1 = bitcast [100 x i32]* %B.i to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %1) nounwind
+  call void @bar([100 x i32]* %A.i, [100 x i32]* %B.i) nounwind
+  call void @llvm.lifetime.end(i64 -1, i8* %0) nounwind
+  call void @llvm.lifetime.end(i64 -1, i8* %1) nounwind
+  %2 = bitcast [100 x i32]* %A.i1 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %2) nounwind
+  %3 = bitcast [100 x i32]* %B.i2 to i8*
+  call void @llvm.lifetime.start(i64 -1, i8* %3) nounwind
+  call void @bar([100 x i32]* %A.i1, [100 x i32]* %B.i2) nounwind
+  call void @llvm.lifetime.end(i64 -1, i8* %2) nounwind
+  call void @llvm.lifetime.end(i64 -1, i8* %3) nounwind
+  ret void
+}
+
+declare void @bar([100 x i32]* , [100 x i32]*) nounwind
+
+declare void @llvm.lifetime.start(i64, i8* nocapture) nounwind
+
+declare void @llvm.lifetime.end(i64, i8* nocapture) nounwind
+
+ declare i32 @foo(i32, i8*)
+

utils/TableGen/CodeGenTarget.cpp

@@ -300,6 +300,8 @@ void CodeGenTarget::ComputeInstrsByEnum() const {
     "REG_SEQUENCE",
     "COPY",
     "BUNDLE",
+    "LIFETIME_START",
+    "LIFETIME_END",
     0
   };
   const DenseMap<const Record*, CodeGenInstruction*> &Insts = getInstructions();