Use 32-bit ebp for NaCl64 in a limited case: llvm.frameaddress.

[jvoung]

Summary

Follow up to [x32] "Use ebp/esp as frame and stack pointer":
http://reviews.llvm.org/D4617

In that earlier patch, NaCl64 was made to always use rbp.
That's needed for most cases because rbp should hold a full
64-bit address within the NaCl sandbox so that load/stores
off of rbp don't require sandbox adjustment (zeroing the top
32-bits, then filling those by adding r15).

However, llvm.frameaddress returns a pointer and pointers
are 32-bit for NaCl64. In this case, use ebp instead, which
will make the register copy type check. A similar mechanism
may be needed for llvm.eh.return, but is not added in this change.

[dschuff]

LGTM

[jvoung]

Closed by commit rL223510 (authored by @jvoung).

lib/Target/X86/X86FastISel.cpp

case Intrinsic::frameaddress: {
     const TargetRegisterClass *RC = nullptr;
 
     switch (VT.SimpleTy) {
     default: llvm_unreachable("Invalid result type for frameaddress.");
     case MVT::i32: Opc = X86::MOV32rm; RC = &X86::GR32RegClass; break;
     case MVT::i64: Opc = X86::MOV64rm; RC = &X86::GR64RegClass; break;
     }
 
-    // This needs to be set before we call getFrameRegister, otherwise we get
-    // the wrong frame register.
+    // This needs to be set before we call getPtrSizedFrameRegister, otherwise
+    // we get the wrong frame register.
     MachineFrameInfo *MFI = FuncInfo.MF->getFrameInfo();
     MFI->setFrameAddressIsTaken(true);
 
     const X86RegisterInfo *RegInfo = static_cast<const X86RegisterInfo *>(
         TM.getSubtargetImpl()->getRegisterInfo());
-    unsigned FrameReg = RegInfo->getFrameRegister(*(FuncInfo.MF));
+    unsigned FrameReg = RegInfo->getPtrSizedFrameRegister(*(FuncInfo.MF));
     assert(((FrameReg == X86::RBP && VT == MVT::i64) ||
             (FrameReg == X86::EBP && VT == MVT::i32)) &&
            "Invalid Frame Register!");
 
     // Always make a copy of the frame register to to a vreg first, so that we
     // never directly reference the frame register (the TwoAddressInstruction-
     // Pass doesn't like that).
     unsigned SrcReg = createResultReg(RC);

lib/Target/X86/X86ISelLowering.cpp

SDValue X86TargetLowering::LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const {
   MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
   MFI->setFrameAddressIsTaken(true);
 
   EVT VT = Op.getValueType();
   SDLoc dl(Op);  // FIXME probably not meaningful
   unsigned Depth = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
   const X86RegisterInfo *RegInfo = static_cast<const X86RegisterInfo *>(
       DAG.getSubtarget().getRegisterInfo());
-  unsigned FrameReg = RegInfo->getFrameRegister(DAG.getMachineFunction());
+  unsigned FrameReg = RegInfo->getPtrSizedFrameRegister(
+      DAG.getMachineFunction());
   assert(((FrameReg == X86::RBP && VT == MVT::i64) ||
           (FrameReg == X86::EBP && VT == MVT::i32)) &&
          "Invalid Frame Register!");
   SDValue FrameAddr = DAG.getCopyFromReg(DAG.getEntryNode(), dl, FrameReg, VT);
   while (Depth--)
     FrameAddr = DAG.getLoad(VT, dl, DAG.getEntryNode(), FrameAddr,
                             MachinePointerInfo(),
                             false, false, false, 0);

lib/Target/X86/X86RegisterInfo.h

bool hasReservedSpillSlot(const MachineFunction &MF, unsigned Reg,
                             int &FrameIdx) const override;
 
   void eliminateFrameIndex(MachineBasicBlock::iterator MI,
                            int SPAdj, unsigned FIOperandNum,
                            RegScavenger *RS = nullptr) const override;
 
   // Debug information queries.
   unsigned getFrameRegister(const MachineFunction &MF) const override;
+  unsigned getPtrSizedFrameRegister(const MachineFunction &MF) const;
   unsigned getStackRegister() const { return StackPtr; }
   unsigned getBaseRegister() const { return BasePtr; }
   // FIXME: Move to FrameInfok
   unsigned getSlotSize() const { return SlotSize; }
 };
 
 // getX86SubSuperRegister - X86 utility function. It returns the sub or super
 // register of a specific X86 register.

lib/Target/X86/X86RegisterInfo.cpp

X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
   }
 }
 
 unsigned X86RegisterInfo::getFrameRegister(const MachineFunction &MF) const {
   const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering();
   return TFI->hasFP(MF) ? FramePtr : StackPtr;
 }
 
+unsigned X86RegisterInfo::getPtrSizedFrameRegister(
+    const MachineFunction &MF) const {
+  unsigned FrameReg = getFrameRegister(MF);
+  if (Subtarget.isTarget64BitILP32())
+    FrameReg = getX86SubSuperRegister(FrameReg, MVT::i32, false);
+  return FrameReg;
+}
+
 namespace llvm {
 unsigned getX86SubSuperRegister(unsigned Reg, MVT::SimpleValueType VT,
                                 bool High) {
   switch (VT) {
   default: llvm_unreachable("Unexpected VT");
   case MVT::i8:
     if (High) {
       switch (Reg) {

test/CodeGen/X86/frameaddr.ll

 ; RUN: llc < %s -march=x86                                | FileCheck %s --check-prefix=CHECK-32
 ; RUN: llc < %s -march=x86    -fast-isel -fast-isel-abort | FileCheck %s --check-prefix=CHECK-32
 ; RUN: llc < %s -march=x86-64                             | FileCheck %s --check-prefix=CHECK-64
 ; RUN: llc < %s -march=x86-64 -fast-isel -fast-isel-abort | FileCheck %s --check-prefix=CHECK-64
 ; RUN: llc < %s -mtriple=x86_64-gnux32                    | FileCheck %s --check-prefix=CHECK-X32ABI
 ; RUN: llc < %s -mtriple=x86_64-gnux32 -fast-isel -fast-isel-abort | FileCheck %s --check-prefix=CHECK-X32ABI
+; RUN: llc < %s -mtriple=x86_64-nacl                    | FileCheck %s --check-prefix=CHECK-NACL64
+; RUN: llc < %s -mtriple=x86_64-nacl -fast-isel -fast-isel-abort | FileCheck %s --check-prefix=CHECK-NACL64
 
 define i8* @test1() nounwind {
 entry:
 ; CHECK-32-LABEL: test1
 ; CHECK-32:       push
 ; CHECK-32-NEXT:  movl %esp, %ebp
 ; CHECK-32-NEXT:  movl %ebp, %eax
 ; CHECK-32-NEXT:  pop
 ; CHECK-32-NEXT:  ret
 ; CHECK-64-LABEL: test1
 ; CHECK-64:       push
 ; CHECK-64-NEXT:  movq %rsp, %rbp
 ; CHECK-64-NEXT:  movq %rbp, %rax
 ; CHECK-64-NEXT:  pop
 ; CHECK-64-NEXT:  ret
 ; CHECK-X32ABI-LABEL: test1
 ; CHECK-X32ABI:       pushq %rbp
 ; CHECK-X32ABI-NEXT:  movl %esp, %ebp
 ; CHECK-X32ABI-NEXT:  movl %ebp, %eax
 ; CHECK-X32ABI-NEXT:  popq %rbp
 ; CHECK-X32ABI-NEXT:  ret
+; CHECK-NACL64-LABEL: test1
+; CHECK-NACL64:       pushq %rbp
+; CHECK-NACL64-NEXT:  movq %rsp, %rbp
+; CHECK-NACL64-NEXT:  movl %ebp, %eax
   %0 = tail call i8* @llvm.frameaddress(i32 0)
   ret i8* %0
 }
 
 define i8* @test2() nounwind {
 entry:
 ; CHECK-32-LABEL: test2
 ; CHECK-32:       push

 ; CHECK-64-NEXT:  ret
 ; CHECK-X32ABI-LABEL: test2
 ; CHECK-X32ABI:       pushq %rbp
 ; CHECK-X32ABI-NEXT:  movl %esp, %ebp
 ; CHECK-X32ABI-NEXT:  movl (%ebp), %eax
 ; CHECK-X32ABI-NEXT:  movl (%eax), %eax
 ; CHECK-X32ABI-NEXT:  popq %rbp
 ; CHECK-X32ABI-NEXT:  ret
+; CHECK-NACL64-LABEL: test2
+; CHECK-NACL64:       pushq %rbp
+; CHECK-NACL64-NEXT:  movq %rsp, %rbp
+; CHECK-NACL64-NEXT:  movl (%ebp), %eax
+; CHECK-NACL64-NEXT:  movl (%eax), %eax
   %0 = tail call i8* @llvm.frameaddress(i32 2)
   ret i8* %0
 }
 
 declare i8* @llvm.frameaddress(i32) nounwind readnone