[NFC][CodeGen] Separate TargetRegisterInfo needsStackRealignment and canRealignStack

[tmatheson]

Summary

Currently needsStackRealignment returns false if canRealignStack returns false. This means that the behavior of needsStackRealignment does not correspond to it's name and description; a function might need stack realignment, but if it is not possible then this function returns false. Furthermore, needsStackRealignment is not virtual and therefore some backends have made use of canRealignStack to indicate whether a function needs stack realignment.

This patch attempts to clarify the situation by separating them:

• Make needsStackRealignment virtual so that backends can override it

• Override needsStackRealignment rather than canRealignStack in SIRegisterInfo

• Update some call sites of needsStackRealignment (where backend tests were failing). Some of these may not be necessary.

This change will make it easier in a future change to handle the case where we need to realign the stack but can't do so (for example when the register allocator creates an aligned spill after the frame pointer has been eliminated).

[tmatheson]

SparcFrameLowering: fix the check for failure to realign that was causing the test failure, and remove the old comment explaining needStackRealignment behavior.

[tmatheson]

Ping

[arsenm]

llvm/include/llvm/CodeGen/TargetRegisterInfo.h

917

I've found this pair of functions confusing before. I think the naming and having to call both in most situations is still confusing. I'm also not sure why this is in TargetRegisterInfo, and not TargetFrameLowering.

needsStackRealignment feels too strong given that the target can say we're not realigning the stack anyway. How about something like hasImpliedStackRealignment? And a partner function to call both called something like hasStackRealignment?

[tmatheson]

needsStackRealignment feels too strong given that the target can say we're not realigning the stack anyway. How about something like hasImpliedStackRealignment? And a partner function to call both called something like hasStackRealignment? And a partner function to call both called something like hasStackRealignment?

How about the following naming?

• shouldRealignStack - true if there is any reason the stack should be realigned
• canRealignStack - true if we are still able to realign the stack (e.g. we can still reserve/have reserved a frame pointer)
• hasStackRealignment = shouldRealignStack && canRealignStack (not target customisable, for now at least)

Then targets are free to error on shouldRealignStack && !canRealignStack if appropriate, or ignore/work around it.

[arsenm]

In D98716#2652931, @tmatheson wrote:

needsStackRealignment feels too strong given that the target can say we're not realigning the stack anyway. How about something like hasImpliedStackRealignment? And a partner function to call both called something like hasStackRealignment? And a partner function to call both called something like hasStackRealignment?

How about the following naming?

• shouldRealignStack - true if there is any reason the stack should be realigned
• canRealignStack - true if we are still able to realign the stack (e.g. we can still reserve/have reserved a frame pointer)
• hasStackRealignment = shouldRealignStack && canRealignStack (not target customisable, for now at least)

Then targets are free to error on shouldRealignStack && !canRealignStack if appropriate, or ignore/work around it.

Seems reasonable

[tmatheson]

Update with new names: shouldRealignStack/canRealignStack/hasStackRealignment

[tmatheson]

Rebase

[tmatheson]

It looks like the clang-tidy CI is broken (running clang tidy on backends that have been modified but are not built?). The rest of the CI looks ok so I will merge this.

llvm/include/llvm/CodeGen/TargetRegisterInfo.h

virtual bool trackLivenessAfterRegAlloc(const MachineFunction &MF) const {
     return true;
   }
 
   /// True if the stack can be realigned for the target.
   virtual bool canRealignStack(const MachineFunction &MF) const;
 
   /// True if storage within the function requires the stack pointer to be
   /// aligned more than the normal calling convention calls for.
-  /// This cannot be overriden by the target, but canRealignStack can be
-  /// overridden.
-  bool needsStackRealignment(const MachineFunction &MF) const;
+  virtual bool needsStackRealignment(const MachineFunction &MF) const;
 
   /// Get the offset from the referenced frame index in the instruction,
   /// if there is one.
   virtual int64_t getFrameIndexInstrOffset(const MachineInstr *MI,
                                            int Idx) const {
     return 0;
   }
 

llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp

void CodeViewDebug::beginFunctionImpl(const MachineFunction *MF) {
   CurFn->Begin = Asm->getFunctionBegin();
 
   // The S_FRAMEPROC record reports the stack size, and how many bytes of
   // callee-saved registers were used. For targets that don't use a PUSH
   // instruction (AArch64), this will be zero.
   CurFn->CSRSize = MFI.getCVBytesOfCalleeSavedRegisters();
   CurFn->FrameSize = MFI.getStackSize();
   CurFn->OffsetAdjustment = MFI.getOffsetAdjustment();
-  CurFn->HasStackRealignment = TRI->needsStackRealignment(*MF);
+  CurFn->HasStackRealignment =
+      TRI->needsStackRealignment(*MF) && TRI->canRealignStack(*MF);
 
   // For this function S_FRAMEPROC record, figure out which codeview register
   // will be the frame pointer.
   CurFn->EncodedParamFramePtrReg = EncodedFramePtrReg::None; // None.
   CurFn->EncodedLocalFramePtrReg = EncodedFramePtrReg::None; // None.
   if (CurFn->FrameSize > 0) {
     if (!TSI.getFrameLowering()->hasFP(*MF)) {
       CurFn->EncodedLocalFramePtrReg = EncodedFramePtrReg::StackPtr;

llvm/lib/CodeGen/GCRootLowering.cpp

bool GCMachineCodeAnalysis::runOnMachineFunction(MachineFunction &MF) {
 
   FI = &getAnalysis<GCModuleInfo>().getFunctionInfo(MF.getFunction());
   TII = MF.getSubtarget().getInstrInfo();
 
   // Find the size of the stack frame.  There may be no correct static frame
   // size, we use UINT64_MAX to represent this.
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
-  const bool DynamicFrameSize = MFI.hasVarSizedObjects() ||
-    RegInfo->needsStackRealignment(MF);
+  const bool DynamicFrameSize =
+      MFI.hasVarSizedObjects() ||
+      (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF));
   FI->setFrameSize(DynamicFrameSize ? UINT64_MAX : MFI.getStackSize());
 
   // Find all safe points.
   if (FI->getStrategy().needsSafePoints())
     FindSafePoints(MF);
 
   // Find the concrete stack offsets for all roots (stack slots)
   FindStackOffsets(MF);
 
   return false;
 }

llvm/lib/CodeGen/MachineFrameInfo.cpp

uint64_t MachineFrameInfo::estimateStackSize(const MachineFunction &MF) const {
 
   // Round up the size to a multiple of the alignment.  If the function has
   // any calls or alloca's, align to the target's StackAlignment value to
   // ensure that the callee's frame or the alloca data is suitably aligned;
   // otherwise, for leaf functions, align to the TransientStackAlignment
   // value.
   Align StackAlign;
   if (adjustsStack() || hasVarSizedObjects() ||
-      (RegInfo->needsStackRealignment(MF) && getObjectIndexEnd() != 0))
+      (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF) &&
+       getObjectIndexEnd() != 0))
     StackAlign = TFI->getStackAlign();
   else
     StackAlign = TFI->getTransientStackAlign();
 
   // If the frame pointer is eliminated, all frame offsets will be relative to
   // SP not FP. Align to MaxAlign so this works.
   StackAlign = std::max(StackAlign, MaxAlign);
   return alignTo(Offset, StackAlign);

llvm/lib/CodeGen/PrologEpilogInserter.cpp

#endif
   // stack area.
   int64_t FixedCSEnd = Offset;
   Align MaxAlign = MFI.getMaxAlign();
 
   // Make sure the special register scavenging spill slot is closest to the
   // incoming stack pointer if a frame pointer is required and is closer
   // to the incoming rather than the final stack pointer.
   const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
-  bool EarlyScavengingSlots = (TFI.hasFP(MF) &&
-                               TFI.isFPCloseToIncomingSP() &&
-                               RegInfo->useFPForScavengingIndex(MF) &&
-                               !RegInfo->needsStackRealignment(MF));
+  bool EarlyScavengingSlots =
+      (TFI.hasFP(MF) && TFI.isFPCloseToIncomingSP() &&
+       !(RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF)));
   if (RS && EarlyScavengingSlots) {
     SmallVector<int, 2> SFIs;
     RS->getScavengingFrameIndices(SFIs);
     for (int SFI : SFIs)
       AdjustStackOffset(MFI, SFI, StackGrowsDown, Offset, MaxAlign, Skew);
   }
 
   // FIXME: Once this is working, then enable flag will change to a target
if (!TFI.targetHandlesStackFrameRounding()) {
 
     // Round up the size to a multiple of the alignment.  If the function has
     // any calls or alloca's, align to the target's StackAlignment value to
     // ensure that the callee's frame or the alloca data is suitably aligned;
     // otherwise, for leaf functions, align to the TransientStackAlignment
     // value.
     Align StackAlign;
     if (MFI.adjustsStack() || MFI.hasVarSizedObjects() ||
-        (RegInfo->needsStackRealignment(MF) && MFI.getObjectIndexEnd() != 0))
+        (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF) &&
+         MFI.getObjectIndexEnd() != 0))
       StackAlign = TFI.getStackAlign();
     else
       StackAlign = TFI.getTransientStackAlign();
 
     // If the frame pointer is eliminated, all frame offsets will be relative to
     // SP not FP. Align to MaxAlign so this works.
     StackAlign = std::max(StackAlign, MaxAlign);
     int64_t OffsetBeforeAlignment = Offset;

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

static SDValue getMemcpyLoadsAndStores(SelectionDAG &DAG, const SDLoc &dl,
 
   if (DstAlignCanChange) {
     Type *Ty = MemOps[0].getTypeForEVT(C);
     Align NewAlign = DL.getABITypeAlign(Ty);
 
     // Don't promote to an alignment that would require dynamic stack
     // realignment.
     const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
-    if (!TRI->needsStackRealignment(MF))
+    if (!TRI->needsStackRealignment(MF) || !TRI->canRealignStack(MF))
       while (NewAlign > Alignment && DL.exceedsNaturalStackAlignment(NewAlign))
         NewAlign = NewAlign / 2;
 
     if (NewAlign > Alignment) {
       // Give the stack frame object a larger alignment if needed.
       if (MFI.getObjectAlign(FI->getIndex()) < NewAlign)
         MFI.setObjectAlignment(FI->getIndex(), NewAlign);
       Alignment = NewAlign;

llvm/lib/CodeGen/StackMaps.cpp

void StackMaps::recordStackMapOpers(const MCSymbol &MILabel,
 
   CSInfos.emplace_back(CSOffsetExpr, ID, std::move(Locations),
                        std::move(LiveOuts));
 
   // Record the stack size of the current function and update callsite count.
   const MachineFrameInfo &MFI = AP.MF->getFrameInfo();
   const TargetRegisterInfo *RegInfo = AP.MF->getSubtarget().getRegisterInfo();
   bool HasDynamicFrameSize =
-      MFI.hasVarSizedObjects() || RegInfo->needsStackRealignment(*(AP.MF));
+      MFI.hasVarSizedObjects() || (RegInfo->needsStackRealignment(*AP.MF) &&
+                                   RegInfo->canRealignStack(*AP.MF));
   uint64_t FrameSize = HasDynamicFrameSize ? UINT64_MAX : MFI.getStackSize();
 
   auto CurrentIt = FnInfos.find(AP.CurrentFnSym);
   if (CurrentIt != FnInfos.end())
     CurrentIt->second.RecordCount++;
   else
     FnInfos.insert(std::make_pair(AP.CurrentFnSym, FunctionInfo(FrameSize)));
 }

llvm/lib/CodeGen/TargetRegisterInfo.cpp

bool TargetRegisterInfo::canRealignStack(const MachineFunction &MF) const {
   return !MF.getFunction().hasFnAttribute("no-realign-stack");
 }
 
 bool TargetRegisterInfo::needsStackRealignment(
     const MachineFunction &MF) const {
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering();
   const Function &F = MF.getFunction();
-  Align StackAlign = TFI->getStackAlign();
-  bool requiresRealignment = ((MFI.getMaxAlign() > StackAlign) ||
-                              F.hasFnAttribute(Attribute::StackAlignment));
-  if (F.hasFnAttribute("stackrealign") || requiresRealignment) {
-    if (canRealignStack(MF))
-      return true;
-    LLVM_DEBUG(dbgs() << "Can't realign function's stack: " << F.getName()
-                      << "\n");
-  }
-  return false;
+  return F.hasFnAttribute("stackrealign") ||
+         (MFI.getMaxAlign() > TFI->getStackAlign()) ||
+         F.hasFnAttribute(Attribute::StackAlignment);
 }
 
 bool TargetRegisterInfo::regmaskSubsetEqual(const uint32_t *mask0,
                                             const uint32_t *mask1) const {
   unsigned N = (getNumRegs()+31) / 32;
   for (unsigned I = 0; I < N; ++I)
     if ((mask0[I] & mask1[I]) != mask0[I])
       return false;

llvm/lib/Target/AMDGPU/SIRegisterInfo.h

unsigned getCSRFirstUseCost() const override {
     return 100;
   }
 
   Register getFrameRegister(const MachineFunction &MF) const override;
 
   bool hasBasePointer(const MachineFunction &MF) const;
   Register getBaseRegister() const;
 
-  bool canRealignStack(const MachineFunction &MF) const override;
+  bool needsStackRealignment(const MachineFunction &MF) const override;
   bool requiresRegisterScavenging(const MachineFunction &Fn) const override;
 
   bool requiresFrameIndexScavenging(const MachineFunction &MF) const override;
   bool requiresFrameIndexReplacementScavenging(
     const MachineFunction &MF) const override;
   bool requiresVirtualBaseRegisters(const MachineFunction &Fn) const override;
 
   int64_t getScratchInstrOffset(const MachineInstr *MI) const;

llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp

for (MCPhysReg Reg : MFI->getVGPRSpillAGPRs())
     reserveRegisterTuples(Reserved, Reg);
 
   for (auto SSpill : MFI->getSGPRSpillVGPRs())
     reserveRegisterTuples(Reserved, SSpill.VGPR);
 
   return Reserved;
 }
 
-bool SIRegisterInfo::canRealignStack(const MachineFunction &MF) const {
+bool SIRegisterInfo::needsStackRealignment(const MachineFunction &MF) const {
   const SIMachineFunctionInfo *Info = MF.getInfo<SIMachineFunctionInfo>();
   // On entry, the base address is 0, so it can't possibly need any more
   // alignment.
 
   // FIXME: Should be able to specify the entry frame alignment per calling
   // convention instead.
   if (Info->isEntryFunction())
     return false;
 
-  return TargetRegisterInfo::canRealignStack(MF);
+  return TargetRegisterInfo::needsStackRealignment(MF);
 }
 
 bool SIRegisterInfo::requiresRegisterScavenging(const MachineFunction &Fn) const {
   const SIMachineFunctionInfo *Info = Fn.getInfo<SIMachineFunctionInfo>();
   if (Info->isEntryFunction()) {
     const MachineFrameInfo &MFI = Fn.getFrameInfo();
     return MFI.hasStackObjects() || MFI.hasCalls();
   }

llvm/lib/Target/ARM/ARMBaseRegisterInfo.cpp

 bool ARMBaseRegisterInfo::hasBasePointer(const MachineFunction &MF) const {
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
   const ARMFrameLowering *TFI = getFrameLowering(MF);
 
   // If we have stack realignment and VLAs, we have no pointer to use to
   // access the stack. If we have stack realignment, and a large call frame,
   // we have no place to allocate the emergency spill slot.
-  if (needsStackRealignment(MF) && !TFI->hasReservedCallFrame(MF))
+  if (needsStackRealignment(MF) && canRealignStack(MF) &&
+      !TFI->hasReservedCallFrame(MF))
     return true;
 
   // Thumb has trouble with negative offsets from the FP. Thumb2 has a limited
   // negative range for ldr/str (255), and Thumb1 is positive offsets only.
   //
   // It's going to be better to use the SP or Base Pointer instead. When there
   // are variable sized objects, we can't reference off of the SP, so we
   // reserve a Base Pointer.
bool ARMBaseRegisterInfo::canRealignStack(const MachineFunction &MF) const {
   return MRI->canReserveReg(BasePtr);
 }
 
 bool ARMBaseRegisterInfo::
 cannotEliminateFrame(const MachineFunction &MF) const {
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   if (MF.getTarget().Options.DisableFramePointerElim(MF) && MFI.adjustsStack())
     return true;
-  return MFI.hasVarSizedObjects() || MFI.isFrameAddressTaken()
-    || needsStackRealignment(MF);
+  return MFI.hasVarSizedObjects() || MFI.isFrameAddressTaken() ||
+         (needsStackRealignment(MF) && canRealignStack(MF));
 }
 
 Register
 ARMBaseRegisterInfo::getFrameRegister(const MachineFunction &MF) const {
   const ARMSubtarget &STI = MF.getSubtarget<ARMSubtarget>();
   const ARMFrameLowering *TFI = getFrameLowering(MF);
 
   if (TFI->hasFP(MF))

llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp

case ARM::Int_eh_sjlj_dispatchsetup: {
           emitThumbRegPlusImmediate(MBB, MBBI, MI.getDebugLoc(), ARM::R6,
                                     FramePtr, -NumBytes, *TII, RI);
         } else {
           emitARMRegPlusImmediate(MBB, MBBI, MI.getDebugLoc(), ARM::R6,
                                   FramePtr, -NumBytes, ARMCC::AL, 0,
                                   *TII);
         }
         // If there's dynamic realignment, adjust for it.
-        if (RI.needsStackRealignment(MF)) {
+        if (RI.needsStackRealignment(MF) && RI.canRealignStack(MF)) {
           MachineFrameInfo &MFI = MF.getFrameInfo();
           Align MaxAlign = MFI.getMaxAlign();
           assert (!AFI->isThumb1OnlyFunction());
           // Emit bic r6, r6, MaxAlign
           assert(MaxAlign <= Align(256) &&
                  "The BIC instruction cannot encode "
                  "immediates larger than 256 with all lower "
                  "bits set.");
           unsigned bicOpc = AFI->isThumbFunction() ?
             ARM::t2BICri : ARM::BICri;
           BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(bicOpc), ARM::R6)
               .addReg(ARM::R6, RegState::Kill)
               .addImm(MaxAlign.value() - 1)
               .add(predOps(ARMCC::AL))
               .add(condCodeOp());
         }
-
       }
       MI.eraseFromParent();
       return true;
     }
 
     case ARM::MOVsrl_flag:
     case ARM::MOVsra_flag: {
       // These are just fancy MOVs instructions.

llvm/lib/Target/ARM/ARMFrameLowering.cpp

bool ARMFrameLowering::hasFP(const MachineFunction &MF) const {
   const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
   const MachineFrameInfo &MFI = MF.getFrameInfo();
 
   // ABI-required frame pointer.
   if (MF.getTarget().Options.DisableFramePointerElim(MF))
     return true;
 
   // Frame pointer required for use within this function.
-  return (RegInfo->needsStackRealignment(MF) ||
-          MFI.hasVarSizedObjects() ||
-          MFI.isFrameAddressTaken());
+  return (
+      (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF)) ||
+      MFI.hasVarSizedObjects() || MFI.isFrameAddressTaken());
 }
 
 /// hasReservedCallFrame - Under normal circumstances, when a frame pointer is
 /// not required, we reserve argument space for call sites in the function
 /// immediately on entry to the current function.  This eliminates the need for
 /// add/sub sp brackets around call sites.  Returns true if the call frame is
 /// included as part of the stack frame.
 bool ARMFrameLowering::hasReservedCallFrame(const MachineFunction &MF) const {
void ARMFrameLowering::emitPrologue(MachineFunction &MF,
   AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
   AFI->setDPRCalleeSavedGapSize(DPRGapSize);
   AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
 
   // If we need dynamic stack realignment, do it here. Be paranoid and make
   // sure if we also have VLAs, we have a base pointer for frame access.
   // If aligned NEON registers were spilled, the stack has already been
   // realigned.
-  if (!AFI->getNumAlignedDPRCS2Regs() && RegInfo->needsStackRealignment(MF)) {
+  if (!AFI->getNumAlignedDPRCS2Regs() && RegInfo->needsStackRealignment(MF) &&
+      RegInfo->canRealignStack(MF)) {
     Align MaxAlign = MFI.getMaxAlign();
     assert(!AFI->isThumb1OnlyFunction());
     if (!AFI->isThumbFunction()) {
       emitAligningInstructions(MF, AFI, TII, MBB, MBBI, dl, ARM::SP, MaxAlign,
                                false);
     } else {
       // We cannot use sp as source/dest register here, thus we're using r4 to
       // perform the calculations. We're emitting the following sequence:
int ARMFrameLowering::ResolveFrameIndexReference(const MachineFunction &MF,
   Offset += SPAdj;
 
   // SP can move around if there are allocas.  We may also lose track of SP
   // when emergency spilling inside a non-reserved call frame setup.
   bool hasMovingSP = !hasReservedCallFrame(MF);
 
   // When dynamically realigning the stack, use the frame pointer for
   // parameters, and the stack/base pointer for locals.
-  if (RegInfo->needsStackRealignment(MF)) {
+  if (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF)) {
     assert(hasFP(MF) && "dynamic stack realignment without a FP!");
     if (isFixed) {
       FrameReg = RegInfo->getFrameRegister(MF);
       Offset = FPOffset;
     } else if (hasMovingSP) {
       assert(RegInfo->hasBasePointer(MF) &&
              "VLAs and dynamic stack alignment, but missing base pointer!");
       FrameReg = RegInfo->getBaseRegister();
void ARMFrameLowering::determineCalleeSaves(MachineFunction &MF,
   Register FramePtr = RegInfo->getFrameRegister(MF);
 
   // Spill R4 if Thumb2 function requires stack realignment - it will be used as
   // scratch register. Also spill R4 if Thumb2 function has varsized objects,
   // since it's not always possible to restore sp from fp in a single
   // instruction.
   // FIXME: It will be better just to find spare register here.
   if (AFI->isThumb2Function() &&
-      (MFI.hasVarSizedObjects() || RegInfo->needsStackRealignment(MF)))
+      (MFI.hasVarSizedObjects() ||
+       (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF))))
     SavedRegs.set(ARM::R4);
 
   // If a stack probe will be emitted, spill R4 and LR, since they are
   // clobbered by the stack probe call.
   // This estimate should be a safe, conservative estimate. The actual
   // stack probe is enabled based on the size of the local objects;
   // this estimate also includes the varargs store size.
   if (STI.isTargetWindows() &&
       WindowsRequiresStackProbe(MF, MFI.estimateStackSize(MF))) {
     SavedRegs.set(ARM::R4);
     SavedRegs.set(ARM::LR);
   }
 
   if (AFI->isThumb1OnlyFunction()) {
     // Spill LR if Thumb1 function uses variable length argument lists.
     if (AFI->getArgRegsSaveSize() > 0)
       SavedRegs.set(ARM::LR);
 
     // Spill R4 if Thumb1 epilogue has to restore SP from FP or the function
     // requires stack alignment.  We don't know for sure what the stack size
     // will be, but for this, an estimate is good enough. If there anything
     // changes it, it'll be a spill, which implies we've used all the registers
     // and so R4 is already used, so not marking it here will be OK.
     // FIXME: It will be better just to find spare register here.
-    if (MFI.hasVarSizedObjects() || RegInfo->needsStackRealignment(MF) ||
+    if (MFI.hasVarSizedObjects() ||
+        (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF)) ||
         MFI.estimateStackSize(MF) > 508)
       SavedRegs.set(ARM::R4);
   }
 
   // See if we can spill vector registers to aligned stack.
   checkNumAlignedDPRCS2Regs(MF, SavedRegs);
 
   // Spill the BasePtr if it's used.

llvm/lib/Target/Sparc/SparcFrameLowering.cpp

void SparcFrameLowering::emitPrologue(MachineFunction &MF,
   const SparcRegisterInfo &RegInfo =
       *static_cast<const SparcRegisterInfo *>(Subtarget.getRegisterInfo());
   MachineBasicBlock::iterator MBBI = MBB.begin();
   // Debug location must be unknown since the first debug location is used
   // to determine the end of the prologue.
   DebugLoc dl;
   bool NeedsStackRealignment = RegInfo.needsStackRealignment(MF);
 
-  // FIXME: unfortunately, returning false from canRealignStack
-  // actually just causes needsStackRealignment to return false,
-  // rather than reporting an error, as would be sensible. This is
-  // poor, but fixing that bogosity is going to be a large project.
-  // For now, just see if it's lied, and report an error here.
-  if (!NeedsStackRealignment && MFI.getMaxAlign() > getStackAlign())
+  if (NeedsStackRealignment && !RegInfo.canRealignStack(MF))
     report_fatal_error("Function \"" + Twine(MF.getName()) + "\" required "
                        "stack re-alignment, but LLVM couldn't handle it "
                        "(probably because it has a dynamic alloca).");
 
   // Get the number of bytes to allocate from the FrameInfo
   int NumBytes = (int) MFI.getStackSize();
 
   unsigned SAVEri = SP::SAVEri;

llvm/lib/Target/X86/X86FrameLowering.cpp

 /// canSimplifyCallFramePseudos - If there is a reserved call frame, the
 /// call frame pseudos can be simplified.  Having a FP, as in the default
 /// implementation, is not sufficient here since we can't always use it.
 /// Use a more nuanced condition.
 bool
 X86FrameLowering::canSimplifyCallFramePseudos(const MachineFunction &MF) const {
   return hasReservedCallFrame(MF) ||
          MF.getInfo<X86MachineFunctionInfo>()->hasPreallocatedCall() ||
-         (hasFP(MF) && !TRI->needsStackRealignment(MF)) ||
+         (hasFP(MF) &&
+          !(TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF))) ||
          TRI->hasBasePointer(MF);
 }
 
 // needsFrameIndexResolution - Do we need to perform FI resolution for
 // this function. Normally, this is required only when the function
 // has any stack objects. However, FI resolution actually has another job,
 // not apparent from the title - it resolves callframesetup/destroy
 // that were not simplified earlier.
 // So, this is required for x86 functions that have push sequences even
 // when there are no stack objects.
 bool
 X86FrameLowering::needsFrameIndexResolution(const MachineFunction &MF) const {
   return MF.getFrameInfo().hasStackObjects() ||
          MF.getInfo<X86MachineFunctionInfo>()->getHasPushSequences();
 }
 
 /// hasFP - Return true if the specified function should have a dedicated frame
 /// pointer register.  This is true if the function has variable sized allocas
 /// or if frame pointer elimination is disabled.
 bool X86FrameLowering::hasFP(const MachineFunction &MF) const {
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   return (MF.getTarget().Options.DisableFramePointerElim(MF) ||
-          TRI->needsStackRealignment(MF) || MFI.hasVarSizedObjects() ||
-          MFI.isFrameAddressTaken() || MFI.hasOpaqueSPAdjustment() ||
+          (TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF)) ||
+          MFI.hasVarSizedObjects() || MFI.isFrameAddressTaken() ||
+          MFI.hasOpaqueSPAdjustment() ||
           MF.getInfo<X86MachineFunctionInfo>()->getForceFramePointer() ||
           MF.getInfo<X86MachineFunctionInfo>()->hasPreallocatedCall() ||
           MF.callsUnwindInit() || MF.hasEHFunclets() || MF.callsEHReturn() ||
           MFI.hasStackMap() || MFI.hasPatchPoint() ||
           MFI.hasCopyImplyingStackAdjustment());
 }
 
 static unsigned getSUBriOpcode(bool IsLP64, int64_t Imm) {
void X86FrameLowering::emitStackProbeInlineGeneric(
   const X86Subtarget &STI = MF.getSubtarget<X86Subtarget>();
   const X86TargetLowering &TLI = *STI.getTargetLowering();
   assert(!(STI.is64Bit() && STI.isTargetWindowsCoreCLR()) &&
          "different expansion expected for CoreCLR 64 bit");
 
   const uint64_t StackProbeSize = TLI.getStackProbeSize(MF);
   uint64_t ProbeChunk = StackProbeSize * 8;
 
-  uint64_t MaxAlign =
-      TRI->needsStackRealignment(MF) ? calculateMaxStackAlign(MF) : 0;
+  uint64_t MaxAlign = TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF)
+                          ? calculateMaxStackAlign(MF)
+                          : 0;
 
   // Synthesize a loop or unroll it, depending on the number of iterations.
   // BuildStackAlignAND ensures that only MaxAlign % StackProbeSize bits left
   // between the unaligned rsp and current rsp.
   if (Offset > ProbeChunk) {
     emitStackProbeInlineGenericLoop(MF, MBB, MBBI, DL, Offset,
                                     MaxAlign % StackProbeSize);
   } else {
if (Fn.getCallingConv() == CallingConv::X86_INTR && Is64Bit &&
     emitSPUpdate(MBB, MBBI, DL, -8, /*InEpilogue=*/false);
   }
 
   // If this is x86-64 and the Red Zone is not disabled, if we are a leaf
   // function, and use up to 128 bytes of stack space, don't have a frame
   // pointer, calls, or dynamic alloca then we do not need to adjust the
   // stack pointer (we fit in the Red Zone). We also check that we don't
   // push and pop from the stack.
-  if (has128ByteRedZone(MF) && !TRI->needsStackRealignment(MF) &&
+  if (has128ByteRedZone(MF) &&
+      !(TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF)) &&
       !MFI.hasVarSizedObjects() &&             // No dynamic alloca.
       !MFI.adjustsStack() &&                   // No calls.
       !EmitStackProbeCall &&                   // No stack probes.
       !MFI.hasCopyImplyingStackAdjustment() && // Don't push and pop.
       !MF.shouldSplitStack()) {                // Regular stack
     uint64_t MinSize = X86FI->getCalleeSavedFrameSize();
     if (HasFP) MinSize += SlotSize;
     X86FI->setUsesRedZone(MinSize > 0 || StackSize > 0);
if (HasFP) {
     uint64_t FrameSize = StackSize - SlotSize;
     // If required, include space for extra hidden slot for stashing base pointer.
     if (X86FI->getRestoreBasePointer())
       FrameSize += SlotSize;
 
     NumBytes = FrameSize - X86FI->getCalleeSavedFrameSize();
 
     // Callee-saved registers are pushed on stack before the stack is realigned.
-    if (TRI->needsStackRealignment(MF) && !IsWin64Prologue)
+    if (TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF) &&
+        !IsWin64Prologue)
       NumBytes = alignTo(NumBytes, MaxAlign);
 
     // Save EBP/RBP into the appropriate stack slot.
     BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::PUSH64r : X86::PUSH32r))
       .addReg(MachineFramePtr, RegState::Kill)
       .setMIFlag(MachineInstr::FrameSetup);
 
     if (NeedsDwarfCFI) {
void X86FrameLowering::emitPrologue(MachineFunction &MF,
   } else {
     assert(!IsFunclet && "funclets without FPs not yet implemented");
     NumBytes = StackSize - X86FI->getCalleeSavedFrameSize();
   }
 
   // Update the offset adjustment, which is mainly used by codeview to translate
   // from ESP to VFRAME relative local variable offsets.
   if (!IsFunclet) {
-    if (HasFP && TRI->needsStackRealignment(MF))
+    if (HasFP && TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF))
       MFI.setOffsetAdjustment(-NumBytes);
     else
       MFI.setOffsetAdjustment(-StackSize);
   }
 
   // For EH funclets, only allocate enough space for outgoing calls. Save the
   // NumBytes value that we would've used for the parent frame.
   unsigned ParentFrameNumBytes = NumBytes;
if (NeedsWinCFI) {
           .addImm(Reg)
           .setMIFlag(MachineInstr::FrameSetup);
     }
   }
 
   // Realign stack after we pushed callee-saved registers (so that we'll be
   // able to calculate their offsets from the frame pointer).
   // Don't do this for Win64, it needs to realign the stack after the prologue.
-  if (!IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF)) {
+  if (!IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF) &&
+      TRI->canRealignStack(MF)) {
     assert(HasFP && "There should be a frame pointer if stack is realigned.");
     BuildStackAlignAND(MBB, MBBI, DL, StackPtr, MaxAlign);
 
     if (NeedsWinCFI) {
       HasWinCFI = true;
       BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_StackAlign))
           .addImm(MaxAlign)
           .setMIFlag(MachineInstr::FrameSetup);
void X86FrameLowering::emitPrologue(MachineFunction &MF,
   // more than 4K bytes on the stack.  Windows uses __chkstk and cygwin/mingw
   // uses __alloca.  __alloca and the 32-bit version of __chkstk will probe the
   // stack and adjust the stack pointer in one go.  The 64-bit version of
   // __chkstk is only responsible for probing the stack.  The 64-bit prologue is
   // responsible for adjusting the stack pointer.  Touching the stack at 4K
   // increments is necessary to ensure that the guard pages used by the OS
   // virtual memory manager are allocated in correct sequence.
   uint64_t AlignedNumBytes = NumBytes;
-  if (IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF))
+  if (IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF) &&
+      TRI->canRealignStack(MF))
     AlignedNumBytes = alignTo(AlignedNumBytes, MaxAlign);
   if (AlignedNumBytes >= StackProbeSize && EmitStackProbeCall) {
     assert(!X86FI->getUsesRedZone() &&
            "The Red Zone is not accounted for in stack probes");
 
     // Check whether EAX is livein for this block.
     bool isEAXAlive = isEAXLiveIn(MBB);
 
addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::MOV64mr)), StackPtr, false,
         .addMemOperand(MF.getMachineMemOperand(
             PSPInfo, MachineMemOperand::MOStore | MachineMemOperand::MOVolatile,
             SlotSize, Align(SlotSize)));
   }
 
   // Realign stack after we spilled callee-saved registers (so that we'll be
   // able to calculate their offsets from the frame pointer).
   // Win64 requires aligning the stack after the prologue.
-  if (IsWin64Prologue && TRI->needsStackRealignment(MF)) {
+  if (IsWin64Prologue && TRI->needsStackRealignment(MF) &&
+      TRI->canRealignStack(MF)) {
     assert(HasFP && "There should be a frame pointer if stack is realigned.");
     BuildStackAlignAND(MBB, MBBI, DL, SPOrEstablisher, MaxAlign);
   }
 
   // We already dealt with stack realignment and funclets above.
   if (IsFunclet && STI.is32Bit())
     return;
 
if (IsFunclet) {
     NumBytes = getWinEHFuncletFrameSize(MF);
   } else if (HasFP) {
     // Calculate required stack adjustment.
     uint64_t FrameSize = StackSize - SlotSize;
     NumBytes = FrameSize - CSSize;
 
     // Callee-saved registers were pushed on stack before the stack was
     // realigned.
-    if (TRI->needsStackRealignment(MF) && !IsWin64Prologue)
+    if (TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF) &&
+        !IsWin64Prologue)
       NumBytes = alignTo(FrameSize, MaxAlign);
   } else {
     NumBytes = StackSize - CSSize;
   }
   uint64_t SEHStackAllocAmt = NumBytes;
 
   // AfterPop is the position to insert .cfi_restore.
   MachineBasicBlock::iterator AfterPop = MBBI;
void X86FrameLowering::emitEpilogue(MachineFunction &MF,
   // instruction, merge the two instructions.
   if (NumBytes || MFI.hasVarSizedObjects())
     NumBytes += mergeSPUpdates(MBB, MBBI, true);
 
   // If dynamic alloca is used, then reset esp to point to the last callee-saved
   // slot before popping them off! Same applies for the case, when stack was
   // realigned. Don't do this if this was a funclet epilogue, since the funclets
   // will not do realignment or dynamic stack allocation.
-  if ((TRI->needsStackRealignment(MF) || MFI.hasVarSizedObjects()) &&
+  if (((TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF)) ||
+       MFI.hasVarSizedObjects()) &&
       !IsFunclet) {
-    if (TRI->needsStackRealignment(MF))
+    if (TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF))
       MBBI = FirstCSPop;
     unsigned SEHFrameOffset = calculateSetFPREG(SEHStackAllocAmt);
     uint64_t LEAAmount =
         IsWin64Prologue ? SEHStackAllocAmt - SEHFrameOffset : -CSSize;
 
     // There are only two legal forms of epilogue:
     // - add SEHAllocationSize, %rsp
     // - lea SEHAllocationSize(%FramePtr), %rsp
StackOffset X86FrameLowering::getFrameIndexReference(const MachineFunction &MF,
   const MachineFrameInfo &MFI = MF.getFrameInfo();
 
   bool IsFixed = MFI.isFixedObjectIndex(FI);
   // We can't calculate offset from frame pointer if the stack is realigned,
   // so enforce usage of stack/base pointer.  The base pointer is used when we
   // have dynamic allocas in addition to dynamic realignment.
   if (TRI->hasBasePointer(MF))
     FrameReg = IsFixed ? TRI->getFramePtr() : TRI->getBaseRegister();
-  else if (TRI->needsStackRealignment(MF))
+  else if (TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF))
     FrameReg = IsFixed ? TRI->getFramePtr() : TRI->getStackRegister();
   else
     FrameReg = TRI->getFrameRegister(MF);
 
   // Offset will hold the offset from the stack pointer at function entry to the
   // object.
   // We need to factor in additional offsets applied during the prologue to the
   // frame, base, and stack pointer depending on which is used.
if (TRI->hasBasePointer(MF)) {
     assert(HasFP && "VLAs and dynamic stack realign, but no FP?!");
     if (FI < 0) {
       // Skip the saved EBP.
       return StackOffset::getFixed(Offset + SlotSize + FPDelta);
     } else {
       assert(isAligned(MFI.getObjectAlign(FI), -(Offset + StackSize)));
       return StackOffset::getFixed(Offset + StackSize);
     }
-  } else if (TRI->needsStackRealignment(MF)) {
+  } else if (TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF)) {
     if (FI < 0) {
       // Skip the saved EBP.
       return StackOffset::getFixed(Offset + SlotSize + FPDelta);
     } else {
       assert(isAligned(MFI.getObjectAlign(FI), -(Offset + StackSize)));
       return StackOffset::getFixed(Offset + StackSize);
     }
     // FIXME: Support tail calls
X86FrameLowering::getFrameIndexReferencePreferSP(const MachineFunction &MF,
   // to address stack arguments for outgoing calls and nothing else. The "base
   // pointer" points to local variables, and RBP points to fixed objects.
   //
   // In cases 2 and 3, we can only answer for non-fixed stack objects, and the
   // answer we give is relative to the SP after the prologue, and not the
   // SP in the middle of the function.
 
   if (MFI.isFixedObjectIndex(FI) && TRI->needsStackRealignment(MF) &&
-      !STI.isTargetWin64())
+      TRI->canRealignStack(MF) && !STI.isTargetWin64())
     return getFrameIndexReference(MF, FI, FrameReg);
 
   // If !hasReservedCallFrame the function might have SP adjustement in the
   // body.  So, even though the offset is statically known, it depends on where
   // we are in the function.
   if (!IgnoreSPUpdates && !hasReservedCallFrame(MF))
     return getFrameIndexReference(MF, FI, FrameReg);
 
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
   }
 
   return I;
 }
 
 bool X86FrameLowering::canUseAsPrologue(const MachineBasicBlock &MBB) const {
   assert(MBB.getParent() && "Block is not attached to a function!");
   const MachineFunction &MF = *MBB.getParent();
-  return !TRI->needsStackRealignment(MF) || !MBB.isLiveIn(X86::EFLAGS);
+  return !TRI->needsStackRealignment(MF) || !TRI->canRealignStack(MF) ||
+         !MBB.isLiveIn(X86::EFLAGS);
 }
 
 bool X86FrameLowering::canUseAsEpilogue(const MachineBasicBlock &MBB) const {
   assert(MBB.getParent() && "Block is not attached to a function!");
 
   // Win64 has strict requirements in terms of epilogue and we are
   // not taking a chance at messing with them.
   // I.e., unless this block is already an exit block, we can't use
void X86FrameLowering::orderFrameObjects(
   for (auto &Obj : SortingObjects) {
     // All invalid items are sorted at the end, so it's safe to stop.
     if (!Obj.IsValid)
       break;
     ObjectsToAllocate[i++] = Obj.ObjectIndex;
   }
 
   // Flip it if we're accessing off of the FP.
-  if (!TRI->needsStackRealignment(MF) && hasFP(MF))
+  if (!(TRI->needsStackRealignment(MF) && TRI->canRealignStack(MF)) &&
+      hasFP(MF))
     std::reverse(ObjectsToAllocate.begin(), ObjectsToAllocate.end());
 }
 
 
 unsigned X86FrameLowering::getWinEHParentFrameOffset(const MachineFunction &MF) const {
   // RDX, the parent frame pointer, is homed into 16(%rsp) in the prologue.
   unsigned Offset = 16;
   // RBP is immediately pushed.

llvm/lib/Target/X86/X86ISelLowering.cpp

bool X86TargetLowering::IsEligibleForTailCallOptimization(
   }
 
   // Look for obvious safe cases to perform tail call optimization that do not
   // require ABI changes. This is what gcc calls sibcall.
 
   // Can't do sibcall if stack needs to be dynamically re-aligned. PEI needs to
   // emit a special epilogue.
   const X86RegisterInfo *RegInfo = Subtarget.getRegisterInfo();
-  if (RegInfo->needsStackRealignment(MF))
+  if (RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF))
     return false;
 
   // Also avoid sibcall optimization if either caller or callee uses struct
   // return semantics.
   if (isCalleeStructRet || isCallerStructRet)
     return false;
 
   // Do not sibcall optimize vararg calls unless all arguments are passed via
case Intrinsic::localaddress: {
     // Returns one of the stack, base, or frame pointer registers, depending on
     // which is used to reference local variables.
     MachineFunction &MF = DAG.getMachineFunction();
     const X86RegisterInfo *RegInfo = Subtarget.getRegisterInfo();
     unsigned Reg;
     if (RegInfo->hasBasePointer(MF))
       Reg = RegInfo->getBaseRegister();
     else { // Handles the SP or FP case.
-      bool CantUseFP = RegInfo->needsStackRealignment(MF);
+      bool CantUseFP =
+          RegInfo->needsStackRealignment(MF) && RegInfo->canRealignStack(MF);
       if (CantUseFP)
         Reg = RegInfo->getPtrSizedStackRegister(MF);
       else
         Reg = RegInfo->getPtrSizedFrameRegister(MF);
     }
     return DAG.getCopyFromReg(DAG.getEntryNode(), dl, Reg, VT);
   }
 

llvm/lib/Target/X86/X86InstrInfo.cpp

if (SubReg && (MO.isDef() || SubReg == X86::sub_8bit_hi))
       return nullptr;
   }
 
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   unsigned Size = MFI.getObjectSize(FrameIndex);
   Align Alignment = MFI.getObjectAlign(FrameIndex);
   // If the function stack isn't realigned we don't want to fold instructions
   // that need increased alignment.
-  if (!RI.needsStackRealignment(MF))
+  if (!RI.needsStackRealignment(MF) || !RI.canRealignStack(MF))
     Alignment =
         std::min(Alignment, Subtarget.getFrameLowering()->getStackAlign());
   if (Ops.size() == 2 && Ops[0] == 0 && Ops[1] == 1) {
     unsigned NewOpc = 0;
     unsigned RCSize = 0;
     switch (MI.getOpcode()) {
     default: return nullptr;
     case X86::TEST8rr:  NewOpc = X86::CMP8ri; RCSize = 1; break;

llvm/lib/Target/X86/X86RegisterInfo.cpp

bool X86RegisterInfo::hasBasePointer(const MachineFunction &MF) const {
   if (!EnableBasePointer)
     return false;
 
   // When we need stack realignment, we can't address the stack from the frame
   // pointer.  When we have dynamic allocas or stack-adjusting inline asm, we
   // can't address variables from the stack pointer.  MS inline asm can
   // reference locals while also adjusting the stack pointer.  When we can't
   // use both the SP and the FP, we need a separate base pointer register.
-  bool CantUseFP = needsStackRealignment(MF);
+  bool CantUseFP = needsStackRealignment(MF) && canRealignStack(MF);
   return CantUseFP && CantUseSP(MFI);
 }
 
 bool X86RegisterInfo::canRealignStack(const MachineFunction &MF) const {
   if (!TargetRegisterInfo::canRealignStack(MF))
     return false;
 
   const MachineFrameInfo &MFI = MF.getFrameInfo();
bool IsEHFuncletEpilogue = MBBI == MBB.end() ? false
                                                : isFuncletReturnInstr(*MBBI);
   const X86FrameLowering *TFI = getFrameLowering(MF);
   int FrameIndex = MI.getOperand(FIOperandNum).getIndex();
 
   // Determine base register and offset.
   int FIOffset;
   Register BasePtr;
   if (MI.isReturn()) {
-    assert((!needsStackRealignment(MF) ||
+    assert((!needsStackRealignment(MF) || !canRealignStack(MF) ||
             MF.getFrameInfo().isFixedObjectIndex(FrameIndex)) &&
            "Return instruction can only reference SP relative frame objects");
     FIOffset =
         TFI->getFrameIndexReferenceSP(MF, FrameIndex, BasePtr, 0).getFixed();
   } else if (TFI->Is64Bit && (MBB.isEHFuncletEntry() || IsEHFuncletEpilogue)) {
     FIOffset = TFI->getWin64EHFrameIndexRef(MF, FrameIndex, BasePtr);
   } else {
     FIOffset = TFI->getFrameIndexReference(MF, FrameIndex, BasePtr).getFixed();
   }