PPC: Add base-pointer support to builtin setjmp/longjmp

First, this changes the base-pointer implementation to remove an unnecessary
complication (and one that is incompatible with how builtin SjLj is
implemented): instead of using r31 as the base pointer when it is not needed as
a frame pointer, now the base pointer will always be r30 when needed.

Second, we introduce another pseudo register, BP, which is used just like the FP
pseudo register to refer to the base register before we know for certain what
register it will be.

Third, we now save BP into the jmp_buf, and restore r30 from that slot in
longjmp.  If the function that called setjmp did not use a base pointer, then
r30 will be overwritten by the setjmp-calling-function's restore code. FP
restoration (which is restored into r31) works the same way.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186545 91177308-0d34-0410-b5e6-96231b3b80d8

7 files changed, 130 insertions, 52 deletions

diff --git a/lib/Target/PowerPC/PPCFrameLowering.cpp b/lib/Target/PowerPC/PPCFrameLowering.cpp
index 3b57390611..d846365112 100644
--- a/lib/Target/PowerPC/PPCFrameLowering.cpp
+++ b/lib/Target/PowerPC/PPCFrameLowering.cpp
@@ -297,6 +297,12 @@ void PPCFrameLowering::replaceFPWithRealFP(MachineFunction &MF) const {
   unsigned FPReg  = is31 ? PPC::R31 : PPC::R1;
   unsigned FP8Reg = is31 ? PPC::X31 : PPC::X1;
 
+  const PPCRegisterInfo *RegInfo =
+    static_cast<const PPCRegisterInfo*>(MF.getTarget().getRegisterInfo());
+  bool HasBP = RegInfo->hasBasePointer(MF);
+  unsigned BPReg  = HasBP ? (unsigned) PPC::R30 : FPReg;
+  unsigned BP8Reg = HasBP ? (unsigned) PPC::X30 : FPReg;
+
   for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
        BI != BE; ++BI)
     for (MachineBasicBlock::iterator MBBI = BI->end(); MBBI != BI->begin(); ) {
@@ -313,6 +319,13 @@ void PPCFrameLowering::replaceFPWithRealFP(MachineFunction &MF) const {
         case PPC::FP8:
           MO.setReg(FP8Reg);
           break;
+        case PPC::BP:
+          MO.setReg(BPReg);
+          break;
+        case PPC::BP8:
+          MO.setReg(BP8Reg);
+          break;
+
         }
       }
     }
@@ -393,8 +406,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const {
       BPOffset = FFI->getObjectOffset(BPIndex);
     } else {
       BPOffset =
-        PPCFrameLowering::getBasePointerSaveOffset(isPPC64, isDarwinABI,
-                                                   HasFP);
+        PPCFrameLowering::getBasePointerSaveOffset(isPPC64, isDarwinABI);
     }
   }
 
@@ -417,7 +429,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const {
 
     if (HasBP)
       BuildMI(MBB, MBBI, dl, TII.get(PPC::STD))
-        .addReg(HasFP ? PPC::X30 : PPC::X31)
+        .addReg(PPC::X30)
         .addImm(BPOffset)
         .addReg(PPC::X1);
 
@@ -448,7 +460,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const {
       // FIXME: On PPC32 SVR4, FPOffset is negative and access to negative
       // offsets of R1 is not allowed.
       BuildMI(MBB, MBBI, dl, TII.get(PPC::STW))
-        .addReg(HasFP ? PPC::R30 : PPC::R31)
+        .addReg(PPC::R30)
         .addImm(BPOffset)
         .addReg(PPC::R1);
 
@@ -475,8 +487,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const {
 
     if (HasBP) {
       // Save a copy of r1 as the base pointer.
-      BuildMI(MBB, MBBI, dl, TII.get(PPC::OR),
-              HasFP ? PPC::R30 : PPC::R31)
+      BuildMI(MBB, MBBI, dl, TII.get(PPC::OR), PPC::R30)
         .addReg(PPC::R1)
         .addReg(PPC::R1);
     }
@@ -527,8 +538,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const {
   } else {    // PPC64.
     if (HasBP) {
       // Save a copy of r1 as the base pointer.
-      BuildMI(MBB, MBBI, dl, TII.get(PPC::OR8),
-              HasFP ? PPC::X30 : PPC::X31)
+      BuildMI(MBB, MBBI, dl, TII.get(PPC::OR8), PPC::X30)
         .addReg(PPC::X1)
         .addReg(PPC::X1);
     }
@@ -597,8 +607,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const {
     }
 
     if (HasBP) {
-      unsigned Reg = isPPC64 ? (HasFP ? PPC::X30 : PPC::X31) :
-                               (HasFP ? PPC::R30 : PPC::R31);
+      unsigned Reg = isPPC64 ? PPC::X30 : PPC::R30;
       Reg = MRI->getDwarfRegNum(Reg, true);
       MMI.addFrameInst(
           MCCFIInstruction::createOffset(FrameLabel, Reg, BPOffset));
@@ -739,8 +748,7 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF,
       BPOffset = FFI->getObjectOffset(BPIndex);
     } else {
       BPOffset =
-        PPCFrameLowering::getBasePointerSaveOffset(isPPC64, isDarwinABI,
-                                                   HasFP);
+        PPCFrameLowering::getBasePointerSaveOffset(isPPC64, isDarwinABI);
     }
   }
 
@@ -836,7 +844,7 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF,
         .addImm(FPOffset).addReg(PPC::X1);
 
     if (HasBP)
-      BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), HasFP ? PPC::X30 : PPC::X31)
+      BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X30)
         .addImm(BPOffset).addReg(PPC::X1);
 
     if (!MustSaveCRs.empty())
@@ -859,7 +867,7 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF,
           .addImm(FPOffset).addReg(PPC::R1);
 
     if (HasBP)
-      BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), HasFP ? PPC::R30 : PPC::R31)
+      BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R30)
           .addImm(FPOffset).addReg(PPC::R1);
 
     if (MustSaveLR)
@@ -968,7 +976,7 @@ PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
 
   int BPSI = FI->getBasePointerSaveIndex();
   if (!BPSI && RegInfo->hasBasePointer(MF)) {
-    int BPOffset = getBasePointerSaveOffset(isPPC64, isDarwinABI, needsFP(MF));
+    int BPOffset = getBasePointerSaveOffset(isPPC64, isDarwinABI);
     // Allocate the frame index for the base pointer save area.
     BPSI = MFI->CreateFixedObject(isPPC64? 8 : 4, BPOffset, true);
     // Save the result.
diff --git a/lib/Target/PowerPC/PPCFrameLowering.h b/lib/Target/PowerPC/PPCFrameLowering.h
index 9acf129be5..7aab37e188 100644
--- a/lib/Target/PowerPC/PPCFrameLowering.h
+++ b/lib/Target/PowerPC/PPCFrameLowering.h
@@ -96,11 +96,7 @@ public:
 
   /// getBasePointerSaveOffset - Return the previous frame offset to save the
   /// base pointer.
-  static unsigned getBasePointerSaveOffset(bool isPPC64, bool isDarwinABI,
-                                           bool hasFP) {
-    if (!hasFP)
-      return getFramePointerSaveOffset(isPPC64, isDarwinABI);
-
+  static unsigned getBasePointerSaveOffset(bool isPPC64, bool isDarwinABI) {
     if (isDarwinABI)
       return isPPC64 ? -16U : -8U;
 
diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp
index a38201abd5..fd225ccad5 100644
--- a/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -6084,6 +6084,7 @@ PPCTargetLowering::emitEHSjLjSetJmp(MachineInstr *MI,
   // thisMBB:
   const int64_t LabelOffset = 1 * PVT.getStoreSize();
   const int64_t TOCOffset   = 3 * PVT.getStoreSize();
+  const int64_t BPOffset    = 4 * PVT.getStoreSize();
 
   // Prepare IP either in reg.
   const TargetRegisterClass *PtrRC = getRegClassFor(PVT);
@@ -6095,10 +6096,25 @@ PPCTargetLowering::emitEHSjLjSetJmp(MachineInstr *MI,
             .addReg(PPC::X2)
             .addImm(TOCOffset)
             .addReg(BufReg);
-
     MIB.setMemRefs(MMOBegin, MMOEnd);
   }
 
+  // Naked functions never have a base pointer, and so we use r1. For all
+  // other functions, this decision must be delayed until during PEI.
+  unsigned BaseReg;
+  if (MF->getFunction()->getAttributes().hasAttribute(
+          AttributeSet::FunctionIndex, Attribute::Naked))
+    BaseReg = PPCSubTarget.isPPC64() ? PPC::X1 : PPC::R1;
+  else
+    BaseReg = PPCSubTarget.isPPC64() ? PPC::BP8 : PPC::BP;
+
+  MIB = BuildMI(*thisMBB, MI, DL,
+                TII->get(PPCSubTarget.isPPC64() ? PPC::STD : PPC::STW))
+          .addReg(BaseReg)
+          .addImm(BPOffset)
+          .addReg(BufReg);
+  MIB.setMemRefs(MMOBegin, MMOEnd);
+
   // Setup
   MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::BCLalways)).addMBB(mainMBB);
   const PPCRegisterInfo *TRI =
@@ -6170,12 +6186,14 @@ PPCTargetLowering::emitEHSjLjLongJmp(MachineInstr *MI,
   // Since FP is only updated here but NOT referenced, it's treated as GPR.
   unsigned FP  = (PVT == MVT::i64) ? PPC::X31 : PPC::R31;
   unsigned SP  = (PVT == MVT::i64) ? PPC::X1 : PPC::R1;
+  unsigned BP  = (PVT == MVT::i64) ? PPC::X30 : PPC::R30;
 
   MachineInstrBuilder MIB;
 
   const int64_t LabelOffset = 1 * PVT.getStoreSize();
   const int64_t SPOffset    = 2 * PVT.getStoreSize();
   const int64_t TOCOffset   = 3 * PVT.getStoreSize();
+  const int64_t BPOffset    = 4 * PVT.getStoreSize();
 
   unsigned BufReg = MI->getOperand(0).getReg();
 
@@ -6217,8 +6235,17 @@ PPCTargetLowering::emitEHSjLjLongJmp(MachineInstr *MI,
   }
   MIB.setMemRefs(MMOBegin, MMOEnd);
 
-  // FIXME: When we also support base pointers, that register must also be
-  // restored here.
+  // Reload BP
+  if (PVT == MVT::i64) {
+    MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), BP)
+            .addImm(BPOffset)
+            .addReg(BufReg);
+  } else {
+    MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), BP)
+            .addImm(BPOffset)
+            .addReg(BufReg);
+  }
+  MIB.setMemRefs(MMOBegin, MMOEnd);
 
   // Reload TOC
   if (PVT == MVT::i64 && PPCSubTarget.isSVR4ABI()) {
diff --git a/lib/Target/PowerPC/PPCRegisterInfo.cpp b/lib/Target/PowerPC/PPCRegisterInfo.cpp
index 49de8da948..fdc604a845 100644
--- a/lib/Target/PowerPC/PPCRegisterInfo.cpp
+++ b/lib/Target/PowerPC/PPCRegisterInfo.cpp
@@ -152,6 +152,11 @@ BitVector PPCRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
   Reserved.set(PPC::FP);
   Reserved.set(PPC::FP8);
 
+  // The BP register is also not really a register, but is the representation
+  // of the base pointer register used by setjmp.
+  Reserved.set(PPC::BP);
+  Reserved.set(PPC::BP8);
+
   // The counter registers must be reserved so that counter-based loops can
   // be correctly formed (and the mtctr instructions are not DCE'd).
   Reserved.set(PPC::CTR);
@@ -178,14 +183,11 @@ BitVector PPCRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
     Reserved.set(PPC::X1);
     Reserved.set(PPC::X13);
 
-    if (PPCFI->needsFP(MF) || hasBasePointer(MF)) {
+    if (PPCFI->needsFP(MF))
       Reserved.set(PPC::X31);
 
-      // If we need a base pointer, and we also have a frame pointer, then use
-      // r30 as the base pointer.
-      if (PPCFI->needsFP(MF) && hasBasePointer(MF))
-        Reserved.set(PPC::X30);
-    }
+    if (hasBasePointer(MF))
+      Reserved.set(PPC::X30);
 
     // The 64-bit SVR4 ABI reserves r2 for the TOC pointer.
     if (Subtarget.isSVR4ABI()) {
@@ -193,12 +195,11 @@ BitVector PPCRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
     }
   }
 
-  if (PPCFI->needsFP(MF) || hasBasePointer(MF)) {
+  if (PPCFI->needsFP(MF))
     Reserved.set(PPC::R31);
 
-    if (PPCFI->needsFP(MF) && hasBasePointer(MF))
-      Reserved.set(PPC::R30);
-  }
+  if (hasBasePointer(MF))
+    Reserved.set(PPC::R30);
 
   // Reserve Altivec registers when Altivec is unavailable.
   if (!Subtarget.hasAltivec())
@@ -675,15 +676,10 @@ unsigned PPCRegisterInfo::getEHHandlerRegister() const {
 }
 
 unsigned PPCRegisterInfo::getBaseRegister(const MachineFunction &MF) const {
-  const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
-
   if (!hasBasePointer(MF))
     return getFrameRegister(MF);
 
-  if (!Subtarget.isPPC64())
-    return TFI->hasFP(MF) ? PPC::R30 : PPC::R31;
-  else
-    return TFI->hasFP(MF) ? PPC::X30 : PPC::X31;
+  return Subtarget.isPPC64() ? PPC::X30 : PPC::R30;
 }
 
 bool PPCRegisterInfo::hasBasePointer(const MachineFunction &MF) const {
diff --git a/lib/Target/PowerPC/PPCRegisterInfo.td b/lib/Target/PowerPC/PPCRegisterInfo.td
index 003e7c3562..d566e2c3e5 100644
--- a/lib/Target/PowerPC/PPCRegisterInfo.td
+++ b/lib/Target/PowerPC/PPCRegisterInfo.td
@@ -94,6 +94,10 @@ def ZERO8 : GP8<ZERO, "0">;
 def FP   : GPR<0 /* arbitrary */, "**FRAME POINTER**">;
 def FP8  : GP8<FP, "**FRAME POINTER**">;
 
+// Representations of the base pointer used by setjmp.
+def BP   : GPR<0 /* arbitrary */, "**BASE POINTER**">;
+def BP8  : GP8<BP, "**BASE POINTER**">;
+
 // Condition register bits
 def CR0LT : CRBIT< 0, "0">;
 def CR0GT : CRBIT< 1, "1">;
@@ -172,11 +176,11 @@ def RM: SPR<512, "**ROUNDING MODE**">;
 // then nonvolatiles in reverse order since stmw/lmw save from rN to r31
 def GPRC : RegisterClass<"PPC", [i32], 32, (add (sequence "R%u", 2, 12),
                                                 (sequence "R%u", 30, 13),
-                                                R31, R0, R1, FP)>;
+                                                R31, R0, R1, FP, BP)>;
 
 def G8RC : RegisterClass<"PPC", [i64], 64, (add (sequence "X%u", 2, 12),
                                                 (sequence "X%u", 30, 14),
-                                                X31, X13, X0, X1, FP8)>;
+                                                X31, X13, X0, X1, FP8, BP8)>;
 
 // For some instructions r0 is special (representing the value 0 instead of
 // the value in the r0 register), and we use these register subclasses to
diff --git a/test/CodeGen/PowerPC/sjlj.ll b/test/CodeGen/PowerPC/sjlj.ll
index 7ea35dafc3..571f3b2fdf 100644
--- a/test/CodeGen/PowerPC/sjlj.ll
+++ b/test/CodeGen/PowerPC/sjlj.ll
@@ -20,6 +20,7 @@ entry:
 ; CHECK: ld [[REG2:[0-9]+]], 8([[REG]])
 ; CHECK: ld 1, 16([[REG]])
 ; CHECK: mtctr [[REG2]]
+; CHECK: ld 30, 32([[REG]])
 ; CHECK: ld 2, 24([[REG]])
 ; CHECK: bctr
 
@@ -99,6 +100,52 @@ return:                                           ; preds = %if.end, %if.then
 ; CHECK-NOAV: blr
 }
 
+define signext i32 @main2() #0 {
+entry:
+  %a = alloca i8, align 64
+  call void @bar(i8* %a)
+  %retval = alloca i32, align 4
+  store i32 0, i32* %retval
+  %0 = call i8* @llvm.frameaddress(i32 0)
+  store i8* %0, i8** bitcast ([1 x %struct.__jmp_buf_tag]* @env_sigill to i8**)
+  %1 = call i8* @llvm.stacksave()
+  store i8* %1, i8** getelementptr (i8** bitcast ([1 x %struct.__jmp_buf_tag]* @env_sigill to i8**), i32 2)
+  %2 = call i32 @llvm.eh.sjlj.setjmp(i8* bitcast ([1 x %struct.__jmp_buf_tag]* @env_sigill to i8*))
+  %tobool = icmp ne i32 %2, 0
+  br i1 %tobool, label %if.then, label %if.else
+
+if.then:                                          ; preds = %entry
+  store i32 1, i32* %retval
+  br label %return
+
+if.else:                                          ; preds = %entry
+  call void @foo()
+  br label %if.end
+
+if.end:                                           ; preds = %if.else
+  store i32 0, i32* %retval
+  br label %return
+
+return:                                           ; preds = %if.end, %if.then
+  %3 = load i32* %retval
+  ret i32 %3
+
+; CHECK: @main2
+
+; CHECK: addis [[REG:[0-9]+]], 2, env_sigill@toc@ha
+; CHECK: std 31, env_sigill@toc@l([[REG]])
+; CHECK: addi [[REG]], [[REG]], env_sigill@toc@l
+; CHECK: std [[REG]], [[OFF:[0-9]+]](31)                  # 8-byte Folded Spill
+; CHECK: std 1, 16([[REG]])
+; CHECK: std 2, 24([[REG]])
+; CHECK: std 30, 32([[REG]])
+; CHECK: bcl 20, 31,
+
+; CHECK: blr
+}
+
+declare void @bar(i8*) #3
+
 declare i8* @llvm.frameaddress(i32) #2
 
 declare i8* @llvm.stacksave() #3
diff --git a/test/CodeGen/PowerPC/stack-realign.ll b/test/CodeGen/PowerPC/stack-realign.ll
index 7bd28f6a7c..f7b6d192c7 100644
--- a/test/CodeGen/PowerPC/stack-realign.ll
+++ b/test/CodeGen/PowerPC/stack-realign.ll
@@ -26,20 +26,20 @@ entry:
 
 ; CHECK-DAG: mflr 0
 ; CHECK-DAG: rldicl [[REG:[0-9]+]], 1, 0, 59
-; CHECK-DAG: std 31, -8(1)
-; CHECK-DAG: mr 31, 1
+; CHECK-DAG: std 30, -16(1)
+; CHECK-DAG: mr 30, 1
 ; CHECK-DAG: std 0, 16(1)
 ; CHECK-DAG: subfic 0, [[REG]], -160
 ; CHECK: stdux 1, 1, 0
 
-; CHECK: .cfi_offset r31, -8
+; CHECK: .cfi_offset r30, -16
 ; CHECK: .cfi_offset lr, 16
 
-; CHECK: std 3, 48(31)
+; CHECK: std 3, 48(30)
 
 ; CHECK: ld 1, 0(1)
 ; CHECK-DAG: ld 0, 16(1)
-; CHECK-DAG: ld 31, -8(1)
+; CHECK-DAG: ld 30, -16(1)
 ; CHECK-DAG: mtlr 0
 ; CHECK: blr
 
@@ -91,8 +91,8 @@ entry:
 ; CHECK-DAG: rldicl [[REG3:[0-9]+]], 1, 0, 59
 ; CHECK-DAG: mflr 0
 ; CHECK-DAG: ori [[REG2:[0-9]+]], [[REG1]], 51808
-; CHECK-DAG: std 31, -8(1)
-; CHECK-DAG: mr 31, 1
+; CHECK-DAG: std 30, -16(1)
+; CHECK-DAG: mr 30, 1
 ; CHECK-DAG: std 0, 16(1)
 ; CHECK-DAG: subfc 0, [[REG3]], [[REG2]]
 ; CHECK: stdux 1, 1, 0
@@ -121,13 +121,13 @@ entry:
 
 ; CHECK-DAG: mflr 0
 ; CHECK-DAG: rldicl [[REG:[0-9]+]], 1, 0, 59
-; CHECK-DAG: std 31, -24(1)
-; CHECK-DAG: mr 31, 1
+; CHECK-DAG: std 30, -32(1)
+; CHECK-DAG: mr 30, 1
 ; CHECK-DAG: std 0, 16(1)
-; CHECK-DAG: subfic 0, [[REG]], -160
+; CHECK-DAG: subfic 0, [[REG]], -192
 ; CHECK: stdux 1, 1, 0
 
-; CHECK: stfd 30, -16(31)
+; CHECK: stfd 30, -16(30)
 
 ; CHECK: blr