[Hexagon] Shrink-wrap stack frame (Hexagon-specific)

llvm-svn: 235603

llvm/lib/Target/Hexagon/HexagonFrameLowering.h

@@ -16,45 +16,42 @@
 namespace llvm {
 
 class HexagonInstrInfo;
+class HexagonRegisterInfo;
 
 class HexagonFrameLowering : public TargetFrameLowering {
-private:
-  void expandAlloca(MachineInstr *AI, const HexagonInstrInfo &TII,
-                    unsigned SP, unsigned CF) const;
-
 public:
   explicit HexagonFrameLowering()
       : TargetFrameLowering(StackGrowsDown, 8, 0, 1, true) {}
 
+  // All of the prolog/epilog functionality, including saving and restoring
+  // callee-saved registers is handled in emitPrologue. This is to have the
+  // logic for shrink-wrapping in one place.
   void emitPrologue(MachineFunction &MF) const override;
-  void emitEpilogue(MachineFunction &MF, MachineBasicBlock &MBB) const override;
+  void emitEpilogue(MachineFunction &MF, MachineBasicBlock &MBB) const
+      override {}
+  bool spillCalleeSavedRegisters(MachineBasicBlock &MBB,
+      MachineBasicBlock::iterator MI, const std::vector<CalleeSavedInfo> &CSI,
+      const TargetRegisterInfo *TRI) const override {
+    return true;
+  }
+  bool restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
+      MachineBasicBlock::iterator MI, const std::vector<CalleeSavedInfo> &CSI,
+      const TargetRegisterInfo *TRI) const override {
+    return true;
+  }
+
+  void eliminateCallFramePseudoInstr(MachineFunction &MF,
+      MachineBasicBlock &MBB, MachineBasicBlock::iterator I) const override;
+  void processFunctionBeforeFrameFinalized(MachineFunction &MF,
+        RegScavenger *RS = nullptr) const override;
+  void processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
+        RegScavenger *RS) const override;
+
   bool targetHandlesStackFrameRounding() const override {
     return true;
   }
-  bool spillCalleeSavedRegisters(MachineBasicBlock &MBB,
-                                 MachineBasicBlock::iterator MI,
-                                 const std::vector<CalleeSavedInfo> &CSI,
-                                 const TargetRegisterInfo *TRI) const override;
-
-  void
-  eliminateCallFramePseudoInstr(MachineFunction &MF,
-                                MachineBasicBlock &MBB,
-                                MachineBasicBlock::iterator I) const override;
-
-  bool
-  restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
-                              MachineBasicBlock::iterator MI,
-                              const std::vector<CalleeSavedInfo> &CSI,
-                              const TargetRegisterInfo *TRI) const override;
-  void processFunctionBeforeFrameFinalized(MachineFunction &MF,
-        RegScavenger *RS = NULL) const override;
-  void processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
-                                            RegScavenger *RS) const override;
   int getFrameIndexOffset(const MachineFunction &MF, int FI) const override;
   bool hasFP(const MachineFunction &MF) const override;
-  bool hasTailCall(MachineBasicBlock &MBB) const;
-  void adjustForCalleeSavedRegsSpillCall(MachineFunction &MF) const;
-  bool replacePredRegPseudoSpillCode(MachineFunction &MF) const;
 
   const SpillSlot *getCalleeSavedSpillSlots(unsigned &NumEntries)
         const override {
@@ -66,17 +63,39 @@ public:
       { Hexagon::R25, -36 }, { Hexagon::R24, -40 }, { Hexagon::D12, -40 },
       { Hexagon::R27, -44 }, { Hexagon::R26, -48 }, { Hexagon::D13, -48 }
     };
-
     NumEntries = array_lengthof(Offsets);
     return Offsets;
   }
 
   bool assignCalleeSavedSpillSlots(MachineFunction &MF,
-        const TargetRegisterInfo *TRI,
+      const TargetRegisterInfo *TRI, std::vector<CalleeSavedInfo> &CSI)
-        std::vector<CalleeSavedInfo> &CSI) const override;
+      const override;
 
   bool needsAligna(const MachineFunction &MF) const;
   MachineInstr *getAlignaInstr(MachineFunction &MF) const;
+
+private:
+  typedef std::vector<CalleeSavedInfo> CSIVect;
+
+  void expandAlloca(MachineInstr *AI, const HexagonInstrInfo &TII,
+      unsigned SP, unsigned CF) const;
+  void insertPrologueInBlock(MachineBasicBlock &MBB) const;
+  void insertEpilogueInBlock(MachineBasicBlock &MBB) const;
+  bool insertCSRSpillsInBlock(MachineBasicBlock &MBB, const CSIVect &CSI,
+      const HexagonRegisterInfo &HRI) const;
+  bool insertCSRRestoresInBlock(MachineBasicBlock &MBB, const CSIVect &CSI,
+      const HexagonRegisterInfo &HRI) const;
+
+  void adjustForCalleeSavedRegsSpillCall(MachineFunction &MF) const;
+  bool replacePredRegPseudoSpillCode(MachineFunction &MF) const;
+  bool replaceVecPredRegPseudoSpillCode(MachineFunction &MF) const;
+
+  void findShrunkPrologEpilog(MachineFunction &MF, MachineBasicBlock *&PrologB,
+      MachineBasicBlock *&EpilogB) const;
+
+  bool shouldInlineCSR(llvm::MachineFunction&, const CSIVect&) const;
+  bool useSpillFunction(MachineFunction &MF, const CSIVect &CSI) const;
+  bool useRestoreFunction(MachineFunction &MF, const CSIVect &CSI) const;
 };
 
 } // End llvm namespace

llvm/lib/Target/Hexagon/HexagonRegisterInfo.h

@@ -37,11 +37,13 @@
 #define HEXAGON_RESERVED_REG_2 Hexagon::R11
 
 namespace llvm {
-struct HexagonRegisterInfo : public HexagonGenRegisterInfo {
+class HexagonRegisterInfo : public HexagonGenRegisterInfo {
+public:
   HexagonRegisterInfo();
 
   /// Code Generation virtual methods...
-  const MCPhysReg *getCalleeSavedRegs(const MachineFunction *MF) const override;
+  const MCPhysReg *getCalleeSavedRegs(const MachineFunction *MF)
+        const override;
 
 
   BitVector getReservedRegs(const MachineFunction &MF) const override;
@@ -76,7 +78,8 @@ struct HexagonRegisterInfo : public HexagonGenRegisterInfo {
   unsigned getFrameRegister() const;
   unsigned getStackRegister() const;
 
-  const uint16_t *getCallerSavedRegs(const MachineFunction *MF) const;
+  const MCPhysReg *getCallerSavedRegs(const MachineFunction *MF) const;
+
   unsigned getFirstCallerSavedNonParamReg() const;
 
   bool isEHReturnCalleeSaveReg(unsigned Reg) const;

llvm/test/CodeGen/Hexagon/shrink-frame-basic.ll

@@ -0,0 +1,36 @@
+; RUN: llc < %s | FileCheck %s
+; Check for allocframe in a non-entry block LBB0_n.
+; CHECK: LBB0_{{[0-9]+}}:
+; CHECK:   allocframe
+; Deallocframe may be in a different block, but must follow.
+; CHECK: deallocframe
+
+target datalayout = "e-m:e-p:32:32-i1:32-i64:64-a:0-v32:32-n16:32"
+target triple = "hexagon"
+
+; Function Attrs: nounwind
+define i32 @foo(i32 %n, i32* %p) #0 {
+entry:
+  %cmp = icmp eq i32* %p, null
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:                                          ; preds = %entry
+  %0 = load i32, i32* %p, align 4
+  %inc = add nsw i32 %0, 1
+  store i32 %inc, i32* %p, align 4
+  br label %return
+
+if.end:                                           ; preds = %entry
+  %call = tail call i32 bitcast (i32 (...)* @bar to i32 (i32)*)(i32 %n) #0
+  %add = add nsw i32 %call, 1
+  br label %return
+
+return:                                           ; preds = %if.end, %if.then
+  %retval.0 = phi i32 [ %0, %if.then ], [ %add, %if.end ]
+  ret i32 %retval.0
+}
+
+declare i32 @bar(...) #0
+
+attributes #0 = { nounwind }
+