1 files changed, 181 insertions, 25 deletions

diff --git a/lib/Target/Mips/MipsSEFrameLowering.cpp b/lib/Target/Mips/MipsSEFrameLowering.cpp
index 33ed4b3..d0a17cd 100644
--- a/lib/Target/Mips/MipsSEFrameLowering.cpp
+++ b/lib/Target/Mips/MipsSEFrameLowering.cpp
@@ -16,6 +16,7 @@
 #include "MipsAnalyzeImmediate.h"
 #include "MipsMachineFunction.h"
 #include "MipsSEInstrInfo.h"
+#include "MipsSubtarget.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
@@ -63,6 +64,10 @@ private:
   bool expandCopy(MachineBasicBlock &MBB, Iter I);
   bool expandCopyACC(MachineBasicBlock &MBB, Iter I, unsigned MFHiOpc,
                      unsigned MFLoOpc);
+  bool expandBuildPairF64(MachineBasicBlock &MBB,
+                          MachineBasicBlock::iterator I, bool FP64) const;
+  bool expandExtractElementF64(MachineBasicBlock &MBB,
+                               MachineBasicBlock::iterator I, bool FP64) const;
 
   MachineFunction &MF;
   MachineRegisterInfo &MRI;
@@ -107,6 +112,22 @@ bool ExpandPseudo::expandInstr(MachineBasicBlock &MBB, Iter I) {
   case Mips::STORE_ACC128:
     expandStoreACC(MBB, I, Mips::PseudoMFHI64, Mips::PseudoMFLO64, 8);
     break;
+  case Mips::BuildPairF64:
+    if (expandBuildPairF64(MBB, I, false))
+      MBB.erase(I);
+    return false;
+  case Mips::BuildPairF64_64:
+    if (expandBuildPairF64(MBB, I, true))
+      MBB.erase(I);
+    return false;
+  case Mips::ExtractElementF64:
+    if (expandExtractElementF64(MBB, I, false))
+      MBB.erase(I);
+    return false;
+  case Mips::ExtractElementF64_64:
+    if (expandExtractElementF64(MBB, I, true))
+      MBB.erase(I);
+    return false;
   case TargetOpcode::COPY:
     if (!expandCopy(MBB, I))
       return false;
@@ -257,6 +278,123 @@ bool ExpandPseudo::expandCopyACC(MachineBasicBlock &MBB, Iter I,
   return true;
 }
 
+/// This method expands the same instruction that MipsSEInstrInfo::
+/// expandBuildPairF64 does, for the case when ABI is fpxx and mthc1 is not
+/// available and the case where the ABI is FP64A. It is implemented here
+/// because frame indexes are eliminated before MipsSEInstrInfo::
+/// expandBuildPairF64 is called.
+bool ExpandPseudo::expandBuildPairF64(MachineBasicBlock &MBB,
+                                      MachineBasicBlock::iterator I,
+                                      bool FP64) const {
+  // For fpxx and when mthc1 is not available, use:
+  //   spill + reload via ldc1
+  //
+  // The case where dmtc1 is available doesn't need to be handled here
+  // because it never creates a BuildPairF64 node.
+  //
+  // The FP64A ABI (fp64 with nooddspreg) must also use a spill/reload sequence
+  // for odd-numbered double precision values (because the lower 32-bits is
+  // transferred with mtc1 which is redirected to the upper half of the even
+  // register). Unfortunately, we have to make this decision before register
+  // allocation so for now we use a spill/reload sequence for all
+  // double-precision values in regardless of being an odd/even register.
+
+  const TargetMachine &TM = MF.getTarget();
+  const MipsSubtarget &Subtarget = TM.getSubtarget<MipsSubtarget>();
+  if ((Subtarget.isABI_FPXX() && !Subtarget.hasMTHC1()) ||
+      (FP64 && !Subtarget.useOddSPReg())) {
+    const MipsSEInstrInfo &TII =
+      *static_cast<const MipsSEInstrInfo*>(TM.getInstrInfo());
+    const MipsRegisterInfo &TRI =
+      *static_cast<const MipsRegisterInfo*>(TM.getRegisterInfo());
+
+    unsigned DstReg = I->getOperand(0).getReg();
+    unsigned LoReg = I->getOperand(1).getReg();
+    unsigned HiReg = I->getOperand(2).getReg();
+
+    // It should be impossible to have FGR64 on MIPS-II or MIPS32r1 (which are
+    // the cases where mthc1 is not available). 64-bit architectures and
+    // MIPS32r2 or later can use FGR64 though.
+    assert(Subtarget.isGP64bit() || Subtarget.hasMTHC1() ||
+           !Subtarget.isFP64bit());
+
+    const TargetRegisterClass *RC = &Mips::GPR32RegClass;
+    const TargetRegisterClass *RC2 =
+        FP64 ? &Mips::FGR64RegClass : &Mips::AFGR64RegClass;
+
+    // We re-use the same spill slot each time so that the stack frame doesn't
+    // grow too much in functions with a large number of moves.
+    int FI = MF.getInfo<MipsFunctionInfo>()->getMoveF64ViaSpillFI(RC2);
+    TII.storeRegToStack(MBB, I, LoReg, I->getOperand(1).isKill(), FI, RC, &TRI,
+                        0);
+    TII.storeRegToStack(MBB, I, HiReg, I->getOperand(2).isKill(), FI, RC, &TRI,
+                        4);
+    TII.loadRegFromStack(MBB, I, DstReg, FI, RC2, &TRI, 0);
+    return true;
+  }
+
+  return false;
+}
+
+/// This method expands the same instruction that MipsSEInstrInfo::
+/// expandExtractElementF64 does, for the case when ABI is fpxx and mfhc1 is not
+/// available and the case where the ABI is FP64A. It is implemented here
+/// because frame indexes are eliminated before MipsSEInstrInfo::
+/// expandExtractElementF64 is called.
+bool ExpandPseudo::expandExtractElementF64(MachineBasicBlock &MBB,
+                                           MachineBasicBlock::iterator I,
+                                           bool FP64) const {
+  // For fpxx and when mfhc1 is not available, use:
+  //   spill + reload via ldc1
+  //
+  // The case where dmfc1 is available doesn't need to be handled here
+  // because it never creates a ExtractElementF64 node.
+  //
+  // The FP64A ABI (fp64 with nooddspreg) must also use a spill/reload sequence
+  // for odd-numbered double precision values (because the lower 32-bits is
+  // transferred with mfc1 which is redirected to the upper half of the even
+  // register). Unfortunately, we have to make this decision before register
+  // allocation so for now we use a spill/reload sequence for all
+  // double-precision values in regardless of being an odd/even register.
+
+  const TargetMachine &TM = MF.getTarget();
+  const MipsSubtarget &Subtarget = TM.getSubtarget<MipsSubtarget>();
+  if ((Subtarget.isABI_FPXX() && !Subtarget.hasMTHC1()) ||
+      (FP64 && !Subtarget.useOddSPReg())) {
+    const MipsSEInstrInfo &TII =
+        *static_cast<const MipsSEInstrInfo *>(TM.getInstrInfo());
+    const MipsRegisterInfo &TRI =
+        *static_cast<const MipsRegisterInfo *>(TM.getRegisterInfo());
+
+    unsigned DstReg = I->getOperand(0).getReg();
+    unsigned SrcReg = I->getOperand(1).getReg();
+    unsigned N = I->getOperand(2).getImm();
+
+    // It should be impossible to have FGR64 on MIPS-II or MIPS32r1 (which are
+    // the cases where mfhc1 is not available). 64-bit architectures and
+    // MIPS32r2 or later can use FGR64 though.
+    assert(Subtarget.isGP64bit() || Subtarget.hasMTHC1() ||
+           !Subtarget.isFP64bit());
+
+    const TargetRegisterClass *RC =
+        FP64 ? &Mips::FGR64RegClass : &Mips::AFGR64RegClass;
+    const TargetRegisterClass *RC2 = &Mips::GPR32RegClass;
+
+    // We re-use the same spill slot each time so that the stack frame doesn't
+    // grow too much in functions with a large number of moves.
+    int FI = MF.getInfo<MipsFunctionInfo>()->getMoveF64ViaSpillFI(RC);
+    TII.storeRegToStack(MBB, I, SrcReg, I->getOperand(1).isKill(), FI, RC, &TRI,
+                        0);
+    TII.loadRegFromStack(MBB, I, DstReg, FI, RC2, &TRI, N * 4);
+    return true;
+  }
+
+  return false;
+}
+
+MipsSEFrameLowering::MipsSEFrameLowering(const MipsSubtarget &STI)
+    : MipsFrameLowering(STI, STI.stackAlignment()) {}
+
 unsigned MipsSEFrameLowering::ehDataReg(unsigned I) const {
   static const unsigned EhDataReg[] = {
     Mips::A0, Mips::A1, Mips::A2, Mips::A3
@@ -299,11 +437,10 @@ void MipsSEFrameLowering::emitPrologue(MachineFunction &MF) const {
   TII.adjustStackPtr(SP, -StackSize, MBB, MBBI);
 
   // emit ".cfi_def_cfa_offset StackSize"
-  MCSymbol *AdjustSPLabel = MMI.getContext().CreateTempSymbol();
-  BuildMI(MBB, MBBI, dl,
-          TII.get(TargetOpcode::PROLOG_LABEL)).addSym(AdjustSPLabel);
-  MMI.addFrameInst(
-      MCCFIInstruction::createDefCfaOffset(AdjustSPLabel, -StackSize));
+  unsigned CFIIndex = MMI.addFrameInst(
+      MCCFIInstruction::createDefCfaOffset(nullptr, -StackSize));
+  BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+      .addCFIIndex(CFIIndex);
 
   const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
 
@@ -315,10 +452,6 @@ void MipsSEFrameLowering::emitPrologue(MachineFunction &MF) const {
 
     // Iterate over list of callee-saved registers and emit .cfi_offset
     // directives.
-    MCSymbol *CSLabel = MMI.getContext().CreateTempSymbol();
-    BuildMI(MBB, MBBI, dl,
-            TII.get(TargetOpcode::PROLOG_LABEL)).addSym(CSLabel);
-
     for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
            E = CSI.end(); I != E; ++I) {
       int64_t Offset = MFI->getObjectOffset(I->getFrameIdx());
@@ -335,14 +468,37 @@ void MipsSEFrameLowering::emitPrologue(MachineFunction &MF) const {
         if (!STI.isLittle())
           std::swap(Reg0, Reg1);
 
-        MMI.addFrameInst(
-            MCCFIInstruction::createOffset(CSLabel, Reg0, Offset));
-        MMI.addFrameInst(
-            MCCFIInstruction::createOffset(CSLabel, Reg1, Offset + 4));
+        unsigned CFIIndex = MMI.addFrameInst(
+            MCCFIInstruction::createOffset(nullptr, Reg0, Offset));
+        BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+            .addCFIIndex(CFIIndex);
+
+        CFIIndex = MMI.addFrameInst(
+            MCCFIInstruction::createOffset(nullptr, Reg1, Offset + 4));
+        BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+            .addCFIIndex(CFIIndex);
+      } else if (Mips::FGR64RegClass.contains(Reg)) {
+        unsigned Reg0 = MRI->getDwarfRegNum(Reg, true);
+        unsigned Reg1 = MRI->getDwarfRegNum(Reg, true) + 1;
+
+        if (!STI.isLittle())
+          std::swap(Reg0, Reg1);
+
+        unsigned CFIIndex = MMI.addFrameInst(
+          MCCFIInstruction::createOffset(nullptr, Reg0, Offset));
+        BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+            .addCFIIndex(CFIIndex);
+
+        CFIIndex = MMI.addFrameInst(
+          MCCFIInstruction::createOffset(nullptr, Reg1, Offset + 4));
+        BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+            .addCFIIndex(CFIIndex);
       } else {
         // Reg is either in GPR32 or FGR32.
-        MMI.addFrameInst(MCCFIInstruction::createOffset(
-            CSLabel, MRI->getDwarfRegNum(Reg, 1), Offset));
+        unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
+            nullptr, MRI->getDwarfRegNum(Reg, 1), Offset));
+        BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+            .addCFIIndex(CFIIndex);
       }
     }
   }
@@ -360,27 +516,27 @@ void MipsSEFrameLowering::emitPrologue(MachineFunction &MF) const {
     }
 
     // Emit .cfi_offset directives for eh data registers.
-    MCSymbol *CSLabel2 = MMI.getContext().CreateTempSymbol();
-    BuildMI(MBB, MBBI, dl,
-            TII.get(TargetOpcode::PROLOG_LABEL)).addSym(CSLabel2);
     for (int I = 0; I < 4; ++I) {
       int64_t Offset = MFI->getObjectOffset(MipsFI->getEhDataRegFI(I));
       unsigned Reg = MRI->getDwarfRegNum(ehDataReg(I), true);
-      MMI.addFrameInst(MCCFIInstruction::createOffset(CSLabel2, Reg, Offset));
+      unsigned CFIIndex = MMI.addFrameInst(
+          MCCFIInstruction::createOffset(nullptr, Reg, Offset));
+      BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+          .addCFIIndex(CFIIndex);
     }
   }
 
   // if framepointer enabled, set it to point to the stack pointer.
   if (hasFP(MF)) {
     // Insert instruction "move $fp, $sp" at this location.
-    BuildMI(MBB, MBBI, dl, TII.get(ADDu), FP).addReg(SP).addReg(ZERO);
+    BuildMI(MBB, MBBI, dl, TII.get(ADDu), FP).addReg(SP).addReg(ZERO)
+      .setMIFlag(MachineInstr::FrameSetup);
 
     // emit ".cfi_def_cfa_register $fp"
-    MCSymbol *SetFPLabel = MMI.getContext().CreateTempSymbol();
-    BuildMI(MBB, MBBI, dl,
-            TII.get(TargetOpcode::PROLOG_LABEL)).addSym(SetFPLabel);
-    MMI.addFrameInst(MCCFIInstruction::createDefCfaRegister(
-        SetFPLabel, MRI->getDwarfRegNum(FP, true)));
+    unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createDefCfaRegister(
+        nullptr, MRI->getDwarfRegNum(FP, true)));
+    BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+        .addCFIIndex(CFIIndex);
   }
 }