Allocate the RS spill slot for any PPC function with spills and a large stack frame

For spills into a large stack frame, the FI-elimination code uses the register
scavenger to obtain a free GPR for use with an r+r-addressed load or store.
When there are no available GPRs, the scavenger gets one by using its spill
slot. Previously, we were not always allocating that spill slot and the RS
would assert when the spill slot was needed.

I don't currently have a small test that triggered the assert, but I've
created a small regression test that verifies that the spill slot is now
added when the stack frame is sufficiently large.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@177140 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/PowerPC/PPCFrameLowering.cpp b/lib/Target/PowerPC/PPCFrameLowering.cpp
index 6feec9f..0f8da5f 100644
--- a/lib/Target/PowerPC/PPCFrameLowering.cpp
+++ b/lib/Target/PowerPC/PPCFrameLowering.cpp
@@ -188,13 +188,21 @@
   return FuncInfo->isCRSpilled();
 }
 
+static bool hasSpills(const MachineFunction &MF) {
+  const PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
+  return FuncInfo->hasSpills();
+}
+
 /// determineFrameLayout - Determine the size of the frame and maximum call
 /// frame size.
-void PPCFrameLowering::determineFrameLayout(MachineFunction &MF) const {
+unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF,
+                                                bool UpdateMF,
+                                                bool UseEstimate) const {
   MachineFrameInfo *MFI = MF.getFrameInfo();
 
   // Get the number of bytes to allocate from the FrameInfo
-  unsigned FrameSize = MFI->getStackSize();
+  unsigned FrameSize =
+    UseEstimate ? MFI->estimateStackSize(MF) : MFI->getStackSize();
 
   // Get the alignments provided by the target, and the maximum alignment
   // (if any) of the fixed frame objects.
@@ -223,8 +231,9 @@
 	&& spillsCR(MF)) &&
       (!ALIGN_STACK || MaxAlign <= TargetAlign)) { // No special alignment.
     // No need for frame
-    MFI->setStackSize(0);
-    return;
+    if (UpdateMF)
+      MFI->setStackSize(0);
+    return 0;
   }
 
   // Get the maximum call frame size of all the calls.
@@ -241,7 +250,8 @@
     maxCallFrameSize = (maxCallFrameSize + AlignMask) & ~AlignMask;
 
   // Update maximum call frame size.
-  MFI->setMaxCallFrameSize(maxCallFrameSize);
+  if (UpdateMF)
+    MFI->setMaxCallFrameSize(maxCallFrameSize);
 
   // Include call frame size in total.
   FrameSize += maxCallFrameSize;
@@ -250,7 +260,10 @@
   FrameSize = (FrameSize + AlignMask) & ~AlignMask;
 
   // Update frame info.
-  MFI->setStackSize(FrameSize);
+  if (UpdateMF)
+    MFI->setStackSize(FrameSize);
+
+  return FrameSize;
 }
 
 // hasFP - Return true if the specified function actually has a dedicated frame
@@ -311,11 +324,7 @@
   MBBI = MBB.begin();
 
   // Work out frame sizes.
-  // FIXME: determineFrameLayout() may change the frame size. This should be
-  // moved upper, to some hook.
-  determineFrameLayout(MF);
-  unsigned FrameSize = MFI->getStackSize();
-
+  unsigned FrameSize = determineFrameLayout(MF);
   int NegFrameSize = -FrameSize;
 
   // Get processor type.
@@ -780,7 +789,7 @@
 
 void
 PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
-                                                   RegScavenger *RS) const {
+                                                   RegScavenger *) const {
   const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo();
 
   //  Save and clear the LR state.
@@ -822,30 +831,15 @@
     int FrameIdx = MFI->CreateFixedObject((uint64_t)4, (int64_t)-4, true);
     FI->setCRSpillFrameIndex(FrameIdx);
   }
-
-  // Reserve a slot closest to SP or frame pointer if we have a dynalloc or
-  // a large stack, which will require scavenging a register to materialize a
-  // large offset.
-  // FIXME: this doesn't actually check stack size, so is a bit pessimistic
-  // FIXME: doesn't detect whether or not we need to spill vXX, which requires
-  //        r0 for now.
-
-  if (RegInfo->requiresRegisterScavenging(MF))
-    if (MFI->hasVarSizedObjects() || spillsCR(MF)) {
-      const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
-      const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
-      const TargetRegisterClass *RC = isPPC64 ? G8RC : GPRC;
-      RS->setScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(),
-                                                         RC->getAlignment(),
-                                                         false));
-    }
 }
 
 void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF,
-                                                         RegScavenger *) const {
+                                                       RegScavenger *RS) const {
   // Early exit if not using the SVR4 ABI.
-  if (!Subtarget.isSVR4ABI())
+  if (!Subtarget.isSVR4ABI()) {
+    addScavengingSpillSlot(MF, RS);
     return;
+  }
 
   // Get callee saved register information.
   MachineFrameInfo *FFI = MF.getFrameInfo();
@@ -853,6 +847,7 @@
 
   // Early exit if no callee saved registers are modified!
   if (CSI.empty() && !needsFP(MF)) {
+    addScavengingSpillSlot(MF, RS);
     return;
   }
 
@@ -1031,6 +1026,37 @@
       FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
     }
   }
+
+  addScavengingSpillSlot(MF, RS);
+}
+
+void
+PPCFrameLowering::addScavengingSpillSlot(MachineFunction &MF,
+                                         RegScavenger *RS) const {
+  // Reserve a slot closest to SP or frame pointer if we have a dynalloc or
+  // a large stack, which will require scavenging a register to materialize a
+  // large offset.
+
+  // We need to have a scavenger spill slot for spills if the frame size is
+  // large. In case there is no free register for large-offset addressing,
+  // this slot is used for the necessary emergency spill. Also, we need the
+  // slot for dynamic stack allocations.
+
+  // The scavenger might be invoked if the frame offset does not fit into
+  // the 16-bit immediate. We don't know the complete frame size here
+  // because we've not yet computed callee-saved register spills or the
+  // needed alignment padding.
+  unsigned StackSize = determineFrameLayout(MF, false, true);
+  MachineFrameInfo *MFI = MF.getFrameInfo();
+  if (MFI->hasVarSizedObjects() || spillsCR(MF) ||
+      (hasSpills(MF) && !isInt<16>(StackSize))) {
+    const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+    const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
+    const TargetRegisterClass *RC = Subtarget.isPPC64() ? G8RC : GPRC;
+    RS->setScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(),
+                                                       RC->getAlignment(),
+                                                       false));
+  }
 }
 
 bool 
diff --git a/lib/Target/PowerPC/PPCFrameLowering.h b/lib/Target/PowerPC/PPCFrameLowering.h
index 8d9d395..53ee326 100644
--- a/lib/Target/PowerPC/PPCFrameLowering.h
+++ b/lib/Target/PowerPC/PPCFrameLowering.h
@@ -32,7 +32,9 @@
       Subtarget(sti) {
   }
 
-  void determineFrameLayout(MachineFunction &MF) const;
+  unsigned determineFrameLayout(MachineFunction &MF,
+                                bool UpdateMF = true,
+                                bool UseEstimate = false) const;
 
   /// emitProlog/emitEpilog - These methods insert prolog and epilog code into
   /// the function.
@@ -46,6 +48,7 @@
                                             RegScavenger *RS = NULL) const;
   void processFunctionBeforeFrameFinalized(MachineFunction &MF,
                                        RegScavenger *RS = NULL) const;
+  void addScavengingSpillSlot(MachineFunction &MF, RegScavenger *RS) const;
 
   bool spillCalleeSavedRegisters(MachineBasicBlock &MBB,
                                  MachineBasicBlock::iterator MI,
diff --git a/lib/Target/PowerPC/PPCInstrInfo.cpp b/lib/Target/PowerPC/PPCInstrInfo.cpp
index 1b4d361..7ae3442 100644
--- a/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ b/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -554,10 +554,11 @@
   MachineFunction &MF = *MBB.getParent();
   SmallVector<MachineInstr*, 4> NewMIs;
 
-  if (StoreRegToStackSlot(MF, SrcReg, isKill, FrameIdx, RC, NewMIs)) {
-    PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
+  PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
+  FuncInfo->setHasSpills();
+
+  if (StoreRegToStackSlot(MF, SrcReg, isKill, FrameIdx, RC, NewMIs))
     FuncInfo->setSpillsCR();
-  }
 
   for (unsigned i = 0, e = NewMIs.size(); i != e; ++i)
     MBB.insert(MI, NewMIs[i]);
diff --git a/lib/Target/PowerPC/PPCMachineFunctionInfo.h b/lib/Target/PowerPC/PPCMachineFunctionInfo.h
index 045b375..10b2160 100644
--- a/lib/Target/PowerPC/PPCMachineFunctionInfo.h
+++ b/lib/Target/PowerPC/PPCMachineFunctionInfo.h
@@ -37,6 +37,9 @@
   /// PEI.
   bool MustSaveLR;
 
+  /// Does this function have any stack spills.
+  bool HasSpills;
+
   /// SpillsCR - Indicates whether CR is spilled in the current function.
   bool SpillsCR;
 
@@ -78,6 +81,7 @@
   explicit PPCFunctionInfo(MachineFunction &MF) 
     : FramePointerSaveIndex(0),
       ReturnAddrSaveIndex(0),
+      HasSpills(false),
       SpillsCR(false),
       LRStoreRequired(false),
       MinReservedArea(0),
@@ -109,6 +113,9 @@
   void setMustSaveLR(bool U) { MustSaveLR = U; }
   bool mustSaveLR() const    { return MustSaveLR; }
 
+  void setHasSpills()      { HasSpills = true; }
+  bool hasSpills() const   { return HasSpills; }
+
   void setSpillsCR()       { SpillsCR = true; }
   bool isCRSpilled() const { return SpillsCR; }