Add support for dynamic stack realignment in the presence of dynamic allocas on

X86.
rdar://11496434


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@158087 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 93 insertions, 14 deletions

diff --git a/lib/Target/X86/X86FrameLowering.cpp b/lib/Target/X86/X86FrameLowering.cpp
index c2b1cf7..244f9bb 100644
--- a/lib/Target/X86/X86FrameLowering.cpp
+++ b/lib/Target/X86/X86FrameLowering.cpp
@@ -650,6 +650,7 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF) const {
   unsigned SlotSize = RegInfo->getSlotSize();
   unsigned FramePtr = RegInfo->getFrameRegister(MF);
   unsigned StackPtr = RegInfo->getStackRegister();
+  unsigned BasePtr = RegInfo->getBaseRegister();
   DebugLoc DL;
 
   // If we're forcing a stack realignment we can't rely on just the frame
@@ -913,6 +914,18 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF) const {
     emitSPUpdate(MBB, MBBI, StackPtr, -(int64_t)NumBytes, Is64Bit,
                  UseLEA, TII, *RegInfo);
 
+  // If we need a base pointer, set it up here. It's whatever the value
+  // of the stack pointer is at this point. Any variable size objects
+  // will be allocated after this, so we can still use the base pointer
+  // to reference locals.
+  if (RegInfo->hasBasePointer(MF)) {
+    // Update the frame pointer with the current stack pointer.
+    unsigned Opc = Is64Bit ? X86::MOV64rr : X86::MOV32rr;
+    BuildMI(MBB, MBBI, DL, TII.get(Opc), BasePtr)
+      .addReg(StackPtr)
+      .setMIFlag(MachineInstr::FrameSetup);
+  }
+
   if (( (!HasFP && NumBytes) || PushedRegs) && needsFrameMoves) {
     // Mark end of stack pointer adjustment.
     MCSymbol *Label = MMI.getContext().CreateTempSymbol();
@@ -1148,7 +1161,16 @@ int X86FrameLowering::getFrameIndexOffset(const MachineFunction &MF, int FI) con
   int Offset = MFI->getObjectOffset(FI) - getOffsetOfLocalArea();
   uint64_t StackSize = MFI->getStackSize();
 
-  if (RegInfo->needsStackRealignment(MF)) {
+  if (RegInfo->hasBasePointer(MF)) {
+    assert (hasFP(MF) && "VLAs and dynamic stack realign, but no FP?!");
+    if (FI < 0) {
+      // Skip the saved EBP.
+      return Offset + RegInfo->getSlotSize();
+    } else {
+      assert((-(Offset + StackSize)) % MFI->getObjectAlignment(FI) == 0);
+      return Offset + StackSize;
+    }
+  } else if (RegInfo->needsStackRealignment(MF)) {
     if (FI < 0) {
       // Skip the saved EBP.
       return Offset + RegInfo->getSlotSize();
@@ -1179,9 +1201,14 @@ int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI,
   const X86RegisterInfo *RegInfo =
       static_cast<const X86RegisterInfo*>(MF.getTarget().getRegisterInfo());
   // We can't calculate offset from frame pointer if the stack is realigned,
-  // so enforce usage of stack pointer.
-  FrameReg = (RegInfo->needsStackRealignment(MF)) ? 
-    RegInfo->getStackRegister() : RegInfo->getFrameRegister(MF);
+  // so enforce usage of stack/base pointer.  The base pointer is used when we
+  // have dynamic allocas in addition to dynamic realignment.
+  if (RegInfo->hasBasePointer(MF))
+    FrameReg = RegInfo->getBaseRegister();
+  else if (RegInfo->needsStackRealignment(MF))
+    FrameReg = RegInfo->getStackRegister();
+  else
+    FrameReg = RegInfo->getFrameRegister(MF);
   return getFrameIndexOffset(MF, FI);
 }
 
@@ -1318,6 +1345,10 @@ X86FrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
            "Slot for EBP register must be last in order to be found!");
     (void)FrameIdx;
   }
+
+  // Spill the BasePtr if it's used.
+  if (RegInfo->hasBasePointer(MF))
+    MF.getRegInfo().setPhysRegUsed(RegInfo->getBaseRegister());
 }
 
 static bool
diff --git a/lib/Target/X86/X86RegisterInfo.cpp b/lib/Target/X86/X86RegisterInfo.cpp
index b22a086..3b72788 100644
--- a/lib/Target/X86/X86RegisterInfo.cpp
+++ b/lib/Target/X86/X86RegisterInfo.cpp
@@ -50,6 +50,10 @@ ForceStackAlign("force-align-stack",
                            " needed for the function."),
                  cl::init(false), cl::Hidden);
 
+cl::opt<bool>
+EnableBasePointer("x86-use-base-pointer", cl::Hidden, cl::init(true),
+          cl::desc("Enable use of a base pointer for complex stack frames"));
+
 X86RegisterInfo::X86RegisterInfo(X86TargetMachine &tm,
                                  const TargetInstrInfo &tii)
   : X86GenRegisterInfo(tm.getSubtarget<X86Subtarget>().is64Bit()
@@ -68,10 +72,12 @@ X86RegisterInfo::X86RegisterInfo(X86TargetMachine &tm,
     SlotSize = 8;
     StackPtr = X86::RSP;
     FramePtr = X86::RBP;
+    BasePtr = X86::RBX;
   } else {
     SlotSize = 4;
     StackPtr = X86::ESP;
     FramePtr = X86::EBP;
+    BasePtr = X86::EBX;
   }
 }
 
@@ -290,6 +296,20 @@ BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const {
       Reserved.set(*I);
   }
 
+  // Set the base-pointer register and its aliases as reserved if needed.
+  if (hasBasePointer(MF)) {
+    CallingConv::ID CC = MF.getFunction()->getCallingConv();
+    const uint32_t* RegMask = getCallPreservedMask(CC);
+    if (MachineOperand::clobbersPhysReg(RegMask, getBaseRegister()))
+      report_fatal_error(
+        "Stack realignment in presence of dynamic allocas is not supported with"
+        "this calling convention.");
+
+    Reserved.set(getBaseRegister());
+    for (MCSubRegIterator I(getBaseRegister(), this); I.isValid(); ++I)
+      Reserved.set(*I);
+  }
+
   // Mark the segment registers as reserved.
   Reserved.set(X86::CS);
   Reserved.set(X86::SS);
@@ -340,10 +360,35 @@ BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const {
 // Stack Frame Processing methods
 //===----------------------------------------------------------------------===//
 
+bool X86RegisterInfo::hasBasePointer(const MachineFunction &MF) const {
+   const MachineFrameInfo *MFI = MF.getFrameInfo();
+
+   if (!EnableBasePointer)
+     return false;
+
+   // When we need stack realignment and there are dynamic allocas, we can't 
+   // reference off of the stack pointer, so we reserve a base pointer.
+   if (needsStackRealignment(MF) && MFI->hasVarSizedObjects())
+     return true;
+
+   return false;
+}
+
 bool X86RegisterInfo::canRealignStack(const MachineFunction &MF) const {
   const MachineFrameInfo *MFI = MF.getFrameInfo();
-  return (MF.getTarget().Options.RealignStack &&
-          !MFI->hasVarSizedObjects());
+  const MachineRegisterInfo *MRI = &MF.getRegInfo();
+  if (!MF.getTarget().Options.RealignStack)
+    return false;
+
+  // Stack realignment requires a frame pointer.  If we already started
+  // register allocation with frame pointer elimination, it is too late now.
+  if (!MRI->canReserveReg(FramePtr))
+    return false;
+
+  // If base pointer is necessary.  Check that it isn't too late to reserve it.
+  if (MFI->hasVarSizedObjects())
+    return MRI->canReserveReg(BasePtr);
+  return true;
 }
 
 bool X86RegisterInfo::needsStackRealignment(const MachineFunction &MF) const {
@@ -353,13 +398,6 @@ bool X86RegisterInfo::needsStackRealignment(const MachineFunction &MF) const {
   bool requiresRealignment = ((MFI->getMaxAlignment() > StackAlign) ||
                                F->hasFnAttr(Attribute::StackAlignment));
 
-  // FIXME: Currently we don't support stack realignment for functions with
-  //        variable-sized allocas.
-  // FIXME: It's more complicated than this...
-  if (0 && requiresRealignment && MFI->hasVarSizedObjects())
-    report_fatal_error(
-      "Stack realignment in presence of dynamic allocas is not supported");
-
   // If we've requested that we force align the stack do so now.
   if (ForceStackAlign)
     return canRealignStack(MF);
@@ -499,7 +537,9 @@ X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
 
   unsigned Opc = MI.getOpcode();
   bool AfterFPPop = Opc == X86::TAILJMPm64 || Opc == X86::TAILJMPm;
-  if (needsStackRealignment(MF))
+  if (hasBasePointer(MF))
+    BasePtr = getBaseRegister();
+  else if (needsStackRealignment(MF))
     BasePtr = (FrameIndex < 0 ? FramePtr : StackPtr);
   else if (AfterFPPop)
     BasePtr = StackPtr;
diff --git a/lib/Target/X86/X86RegisterInfo.h b/lib/Target/X86/X86RegisterInfo.h
index ee69842..1bc32cb 100644
--- a/lib/Target/X86/X86RegisterInfo.h
+++ b/lib/Target/X86/X86RegisterInfo.h
@@ -50,6 +50,11 @@ private:
   ///
   unsigned FramePtr;
 
+  /// BasePtr - X86 physical register used as a base ptr in complex stack
+  /// frames. I.e., when we need a 3rd base, not just SP and FP, due to
+  /// variable size stack objects.
+  unsigned BasePtr;
+
 public:
   X86RegisterInfo(X86TargetMachine &tm, const TargetInstrInfo &tii);
 
@@ -106,6 +111,8 @@ public:
   /// register scavenger to determine what registers are free.
   BitVector getReservedRegs(const MachineFunction &MF) const;
 
+  bool hasBasePointer(const MachineFunction &MF) const;
+
   bool canRealignStack(const MachineFunction &MF) const;
 
   bool needsStackRealignment(const MachineFunction &MF) const;
@@ -123,6 +130,7 @@ public:
   // Debug information queries.
   unsigned getFrameRegister(const MachineFunction &MF) const;
   unsigned getStackRegister() const { return StackPtr; }
+  unsigned getBaseRegister() const { return BasePtr; }
   // FIXME: Move to FrameInfok
   unsigned getSlotSize() const { return SlotSize; }