Rewrite stack callee saved spills and restores to use push/pop instructions.

Remove movePastCSLoadStoreOps and associated code for simple pointer
increments. Update routines that depended upon other opcodes for save/restore.

Adjust all testcases accordingly.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119725 91177308-0d34-0410-b5e6-96231b3b80d8

7 files changed, 193 insertions, 90 deletions

diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 94147eb31e..cb9c11295b 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -197,26 +197,29 @@ ARMBaseInstrInfo::convertToThreeAddress(MachineFunction::iterator &MFI,
   return NewMIs[0];
 }
 
-bool
-ARMBaseInstrInfo::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
-                                        MachineBasicBlock::iterator MI,
-                                        const std::vector<CalleeSavedInfo> &CSI,
-                                        const TargetRegisterInfo *TRI) const {
-  if (CSI.empty())
-    return false;
-
+void
+ARMBaseInstrInfo::emitPushInst(MachineBasicBlock &MBB, 
+                MachineBasicBlock::iterator MI,
+                const std::vector<CalleeSavedInfo> &CSI, unsigned Opc,
+                bool(*Func)(unsigned, bool)) const {
+  MachineFunction &MF = *MBB.getParent();
   DebugLoc DL;
   if (MI != MBB.end()) DL = MI->getDebugLoc();
 
-  for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
-    unsigned Reg = CSI[i].getReg();
-    bool isKill = true;
-
+  MachineInstrBuilder MIB = BuildMI(MF, DL, get(Opc));
+  MIB.addReg(ARM::SP, getDefRegState(true));
+  MIB.addReg(ARM::SP);
+  AddDefaultPred(MIB);
+  bool NumRegs = false;
+  for (unsigned i = CSI.size(); i != 0; --i) {
+    unsigned Reg = CSI[i-1].getReg();
+    if (!(Func)(Reg, Subtarget.isTargetDarwin())) continue;
+    
     // Add the callee-saved register as live-in unless it's LR and
     // @llvm.returnaddress is called. If LR is returned for @llvm.returnaddress
     // then it's already added to the function and entry block live-in sets.
+    bool isKill = true;
     if (Reg == ARM::LR) {
-      MachineFunction &MF = *MBB.getParent();
       if (MF.getFrameInfo()->isReturnAddressTaken() &&
           MF.getRegInfo().isLiveIn(Reg))
         isKill = false;
@@ -224,16 +227,99 @@ ARMBaseInstrInfo::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
 
     if (isKill)
       MBB.addLiveIn(Reg);
+    
+    NumRegs = true;
+    MIB.addReg(Reg, getKillRegState(isKill));
+  }
+  
+  // It's illegal to emit push instruction without operands.
+  if (NumRegs)
+    MBB.insert(MI, &*MIB);
+  else
+    MF.DeleteMachineInstr(MIB);        
+}
+
+bool
+ARMBaseInstrInfo::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
+                                        MachineBasicBlock::iterator MI,
+                                        const std::vector<CalleeSavedInfo> &CSI,
+                                        const TargetRegisterInfo *TRI) const {
+  if (CSI.empty())
+    return false;
+
+  MachineFunction &MF = *MBB.getParent();
+  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
+  DebugLoc DL = MI->getDebugLoc();
+  
+  unsigned PushOpc = AFI->isThumbFunction() ? ARM::t2STMDB_UPD : ARM::STMDB_UPD;
+  unsigned FltOpc = ARM::VSTMDDB_UPD;
+  emitPushInst(MBB, MI, CSI, PushOpc, &isARMArea1Register);
+  emitPushInst(MBB, MI, CSI, PushOpc, &isARMArea2Register);
+  emitPushInst(MBB, MI, CSI, FltOpc, &isARMArea3Register);
+
+  return true;
+}
+
+void
+ARMBaseInstrInfo::emitPopInst(MachineBasicBlock &MBB, 
+                MachineBasicBlock::iterator MI,
+                const std::vector<CalleeSavedInfo> &CSI, unsigned Opc,
+                bool isVarArg, bool(*Func)(unsigned, bool)) const {
+  
+  MachineFunction &MF = *MBB.getParent();
+  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
+  DebugLoc DL = MI->getDebugLoc();
+  
+  MachineInstrBuilder MIB = BuildMI(MF, DL, get(Opc));
+  MIB.addReg(ARM::SP, getDefRegState(true));
+  MIB.addReg(ARM::SP);
+  AddDefaultPred(MIB);
+  bool NumRegs = false;
+  for (unsigned i = CSI.size(); i != 0; --i) {
+    unsigned Reg = CSI[i-1].getReg();
+    if (!(Func)(Reg, Subtarget.isTargetDarwin())) continue;
+
+    if (Reg == ARM::LR && !isVarArg) {
+      Reg = ARM::PC;
+      unsigned Opc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
+      (*MIB).setDesc(get(Opc));
+      MI = MBB.erase(MI);
+    }
 
-    // Insert the spill to the stack frame. The register is killed at the spill
-    //
-    const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
-    storeRegToStackSlot(MBB, MI, Reg, isKill,
-                        CSI[i].getFrameIdx(), RC, TRI);
+    MIB.addReg(Reg, RegState::Define);
+    NumRegs = true;
   }
+    
+  // It's illegal to emit pop instruction without operands.
+  if (NumRegs)
+    MBB.insert(MI, &*MIB);
+  else
+    MF.DeleteMachineInstr(MIB);
+}
+
+bool
+ARMBaseInstrInfo::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
+                                        MachineBasicBlock::iterator MI,
+                                        const std::vector<CalleeSavedInfo> &CSI,
+                                        const TargetRegisterInfo *TRI) const {
+  if (CSI.empty())
+    return false;
+  
+  MachineFunction &MF = *MBB.getParent();
+  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
+  bool isVarArg = AFI->getVarArgsRegSaveSize() > 0;
+  DebugLoc DL = MI->getDebugLoc();
+  
+  unsigned PopOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_UPD : ARM::LDMIA_UPD;
+  unsigned FltOpc = ARM::VLDMDIA_UPD;
+  emitPopInst(MBB, MI, CSI, FltOpc, isVarArg, &isARMArea3Register);
+  emitPopInst(MBB, MI, CSI, PopOpc, isVarArg, &isARMArea2Register);
+  emitPopInst(MBB, MI, CSI, PopOpc, isVarArg, &isARMArea1Register);
+
   return true;
 }
 
+
 // Branch analysis.
 bool
 ARMBaseInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
@@ -2195,7 +2281,7 @@ int ARMBaseInstrInfo::getInstrLatency(const InstrItineraryData *ItinData,
   case ARM::VSTMQIA:
   case ARM::VSTMQDB:
     return 2;
-  }  
+  }
 }
 
 bool ARMBaseInstrInfo::
diff --git a/lib/Target/ARM/ARMBaseInstrInfo.h b/lib/Target/ARM/ARMBaseInstrInfo.h
index cbcc428b71..5e7156c9ce 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.h
+++ b/lib/Target/ARM/ARMBaseInstrInfo.h
@@ -211,6 +211,21 @@ public:
                                  const std::vector<CalleeSavedInfo> &CSI,
                                  const TargetRegisterInfo *TRI) const;
 
+  bool restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
+                                   MachineBasicBlock::iterator MI,
+                                   const std::vector<CalleeSavedInfo> &CSI,
+                                   const TargetRegisterInfo *TRI) const;
+
+private:
+  void emitPopInst(MachineBasicBlock &MBB, MachineBasicBlock::iterator MI,
+                   const std::vector<CalleeSavedInfo> &CSI, unsigned Opc,
+                   bool isVarArg, bool(*Func)(unsigned, bool)) const;
+  void emitPushInst(MachineBasicBlock &MBB, MachineBasicBlock::iterator MI,
+                    const std::vector<CalleeSavedInfo> &CSI, unsigned Opc,
+                    bool(*Func)(unsigned, bool)) const;
+  
+  
+public:
   // Branch analysis.
   virtual bool AnalyzeBranch(MachineBasicBlock &MBB, MachineBasicBlock *&TBB,
                              MachineBasicBlock *&FBB,
diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.h b/lib/Target/ARM/ARMBaseRegisterInfo.h
index 25792c70c3..5cb080804b 100644
--- a/lib/Target/ARM/ARMBaseRegisterInfo.h
+++ b/lib/Target/ARM/ARMBaseRegisterInfo.h
@@ -44,6 +44,45 @@ static inline bool isARMLowRegister(unsigned Reg) {
   }
 }
 
+/// isARMArea1Register - Returns true if the register is a low register (r0-r7)
+/// or a stack/pc register that we should push/pop.
+static inline bool isARMArea1Register(unsigned Reg, bool isDarwin) {
+  using namespace ARM;
+  switch (Reg) {
+    case R0:  case R1:  case R2:  case R3:
+    case R4:  case R5:  case R6:  case R7:
+    case LR:  case SP:  case PC:
+      return true;
+    case R8:  case R9:  case R10: case R11:
+      // For darwin we want r7 and lr to be next to each other.
+      return !isDarwin;
+    default:
+      return false;
+  }
+}
+
+static inline bool isARMArea2Register(unsigned Reg, bool isDarwin) {
+  using namespace ARM;
+  switch (Reg) {
+    case R8: case R9: case R10: case R11:
+      // Darwin has this second area.
+      return isDarwin;
+    default:
+      return false;
+  }
+}
+
+static inline bool isARMArea3Register(unsigned Reg, bool isDarwin) {
+  using namespace ARM;
+  switch (Reg) {
+    case D15: case D14: case D13: case D12:
+    case D11: case D10: case D9:  case D8:
+      return true;
+    default:
+      return false;
+  }
+}
+
 class ARMBaseRegisterInfo : public ARMGenRegisterInfo {
 protected:
   const ARMBaseInstrInfo &TII;
diff --git a/lib/Target/ARM/ARMFrameInfo.cpp b/lib/Target/ARM/ARMFrameInfo.cpp
index 499111af06..3417ae2659 100644
--- a/lib/Target/ARM/ARMFrameInfo.cpp
+++ b/lib/Target/ARM/ARMFrameInfo.cpp
@@ -20,43 +20,6 @@
 
 using namespace llvm;
 
-/// Move iterator past the next bunch of callee save load / store ops for
-/// the particular spill area (1: integer area 1, 2: integer area 2,
-/// 3: fp area, 0: don't care).
-static void movePastCSLoadStoreOps(MachineBasicBlock &MBB,
-                                   MachineBasicBlock::iterator &MBBI,
-                                   int Opc1, int Opc2, unsigned Area,
-                                   const ARMSubtarget &STI) {
-  while (MBBI != MBB.end() &&
-         ((MBBI->getOpcode() == Opc1) || (MBBI->getOpcode() == Opc2)) &&
-         MBBI->getOperand(1).isFI()) {
-    if (Area != 0) {
-      bool Done = false;
-      unsigned Category = 0;
-      switch (MBBI->getOperand(0).getReg()) {
-      case ARM::R4:  case ARM::R5:  case ARM::R6: case ARM::R7:
-      case ARM::LR:
-        Category = 1;
-        break;
-      case ARM::R8:  case ARM::R9:  case ARM::R10: case ARM::R11:
-        Category = STI.isTargetDarwin() ? 2 : 1;
-        break;
-      case ARM::D8:  case ARM::D9:  case ARM::D10: case ARM::D11:
-      case ARM::D12: case ARM::D13: case ARM::D14: case ARM::D15:
-        Category = 3;
-        break;
-      default:
-        Done = true;
-        break;
-      }
-      if (Done || Category != Area)
-        break;
-    }
-
-    ++MBBI;
-  }
-}
-
 static bool isCalleeSavedRegister(unsigned Reg, const unsigned *CSRegs) {
   for (unsigned i = 0; CSRegs[i]; ++i)
     if (Reg == CSRegs[i])
@@ -67,11 +30,21 @@ static bool isCalleeSavedRegister(unsigned Reg, const unsigned *CSRegs) {
 static bool isCSRestore(MachineInstr *MI,
                         const ARMBaseInstrInfo &TII,
                         const unsigned *CSRegs) {
-  return ((MI->getOpcode() == (int)ARM::VLDRD ||
-           MI->getOpcode() == (int)ARM::LDRi12 ||
-           MI->getOpcode() == (int)ARM::t2LDRi12) &&
-          MI->getOperand(1).isFI() &&
-          isCalleeSavedRegister(MI->getOperand(0).getReg(), CSRegs));
+  // Integer spill area is handled with "pop".
+  if (MI->getOpcode() == ARM::LDMIA_RET ||
+      MI->getOpcode() == ARM::t2LDMIA_RET ||
+      MI->getOpcode() == ARM::LDMIA_UPD ||
+      MI->getOpcode() == ARM::t2LDMIA_UPD ||
+      MI->getOpcode() == ARM::VLDMDIA_UPD) {
+    // The first two operands are predicates. The last two are
+    // imp-def and imp-use of SP. Check everything in between.
+    for (int i = 5, e = MI->getNumOperands(); i != e; ++i)
+      if (!isCalleeSavedRegister(MI->getOperand(i).getReg(), CSRegs))
+        return false;
+    return true;
+  }
+
+  return false;
 }
 
 static void
@@ -154,11 +127,10 @@ void ARMFrameInfo::emitPrologue(MachineFunction &MF) const {
       DPRCSSize += 8;
     }
   }
-
-  // Build the new SUBri to adjust SP for integer callee-save spill area 1.
-  emitSPUpdate(isARM, MBB, MBBI, dl, TII, -GPRCS1Size);
-  movePastCSLoadStoreOps(MBB, MBBI, ARM::STRi12, ARM::t2STRi12, 1, STI);
-
+  
+  // Move past area 1.
+  if (GPRCS1Size > 0) MBBI++;
+  
   // Set FP to point to the stack slot that contains the previous FP.
   // For Darwin, FP is R7, which has now been stored in spill area 1.
   // Otherwise, if this is not Darwin, all the callee-saved registers go
@@ -172,14 +144,10 @@ void ARMFrameInfo::emitPrologue(MachineFunction &MF) const {
       .addFrameIndex(FramePtrSpillFI).addImm(0);
     AddDefaultCC(AddDefaultPred(MIB));
   }
-
-  // Build the new SUBri to adjust SP for integer callee-save spill area 2.
-  emitSPUpdate(isARM, MBB, MBBI, dl, TII, -GPRCS2Size);
-
-  // Build the new SUBri to adjust SP for FP callee-save spill area.
-  movePastCSLoadStoreOps(MBB, MBBI, ARM::STRi12, ARM::t2STRi12, 2, STI);
-  emitSPUpdate(isARM, MBB, MBBI, dl, TII, -DPRCSSize);
-
+  
+  // Move past area 2.
+  if (GPRCS2Size > 0) MBBI++;
+  
   // Determine starting offsets of spill areas.
   unsigned DPRCSOffset  = NumBytes - (GPRCS1Size + GPRCS2Size + DPRCSSize);
   unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
@@ -191,7 +159,9 @@ void ARMFrameInfo::emitPrologue(MachineFunction &MF) const {
   AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
   AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
 
-  movePastCSLoadStoreOps(MBB, MBBI, ARM::VSTRD, 0, 3, STI);
+  // Move past area 3.
+  if (DPRCSSize > 0) MBBI++;
+  
   NumBytes = DPRCSOffset;
   if (NumBytes) {
     // Adjust SP after all the callee-save spills.
@@ -325,17 +295,10 @@ void ARMFrameInfo::emitEpilogue(MachineFunction &MF,
     } else if (NumBytes)
       emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);
 
-    // Move SP to start of integer callee save spill area 2.
-    movePastCSLoadStoreOps(MBB, MBBI, ARM::VLDRD, 0, 3, STI);
-    emitSPUpdate(isARM, MBB, MBBI, dl, TII, AFI->getDPRCalleeSavedAreaSize());
-
-    // Move SP to start of integer callee save spill area 1.
-    movePastCSLoadStoreOps(MBB, MBBI, ARM::LDRi12, ARM::t2LDRi12, 2, STI);
-    emitSPUpdate(isARM, MBB, MBBI, dl, TII, AFI->getGPRCalleeSavedArea2Size());
-
-    // Move SP to SP upon entry to the function.
-    movePastCSLoadStoreOps(MBB, MBBI, ARM::LDRi12, ARM::t2LDRi12, 1, STI);
-    emitSPUpdate(isARM, MBB, MBBI, dl, TII, AFI->getGPRCalleeSavedArea1Size());
+    // Increment past our save areas.
+    if (AFI->getDPRCalleeSavedAreaSize()) MBBI++;
+    if (AFI->getGPRCalleeSavedArea2Size()) MBBI++;
+    if (AFI->getGPRCalleeSavedArea1Size()) MBBI++;
   }
 
   if (RetOpcode == ARM::TCRETURNdi || RetOpcode == ARM::TCRETURNdiND ||
diff --git a/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll b/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll
index 2789ccd5cf..163c9b030e 100644
--- a/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll
+++ b/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll
@@ -3,7 +3,7 @@
 ; LDM instruction, was causing an assertion failure because the microop count
 ; was being treated as an instruction count.
 
-; CHECK: ldmia
+; CHECK: push
 ; CHECK: ldmia
 ; CHECK: ldmia
 ; CHECK: ldmia
diff --git a/test/CodeGen/ARM/str_pre-2.ll b/test/CodeGen/ARM/str_pre-2.ll
index 4f9ba4db4c..a79cf9bf7f 100644
--- a/test/CodeGen/ARM/str_pre-2.ll
+++ b/test/CodeGen/ARM/str_pre-2.ll
@@ -4,8 +4,8 @@
 
 define i64 @t(i64 %a) nounwind readonly {
 entry:
-; CHECK: str lr, [sp, #-4]!
-; CHECK: ldr lr, [sp], #4
+; CHECK: push    {lr}
+; CHECK: ldmia   sp!, {pc}
 	%0 = load i64** @b, align 4
 	%1 = load i64* %0, align 4
 	%2 = mul i64 %1, %a
diff --git a/test/CodeGen/Thumb2/2010-06-21-TailMergeBug.ll b/test/CodeGen/Thumb2/2010-06-21-TailMergeBug.ll
index f91e1c9feb..12b4975d69 100644
--- a/test/CodeGen/Thumb2/2010-06-21-TailMergeBug.ll
+++ b/test/CodeGen/Thumb2/2010-06-21-TailMergeBug.ll
@@ -39,7 +39,7 @@ entry:
 ; CHECK: ittt eq
 ; CHECK: moveq r0
 ; CHECK-NOT: LBB0_
-; CHECK: ldreq
+; CHECK: popeq
 ; CHECK: popeq
   switch i32 undef, label %bb7 [
     i32 37, label %bb43