Allow some reloads to be folded in multi-use cases. Specifically testl r, r -> cmpl [mem], 0.

llvm-svn: 44479

2 files changed, 87 insertions, 1 deletions

diff --git a/llvm/lib/Target/X86/X86RegisterInfo.cpp b/llvm/lib/Target/X86/X86RegisterInfo.cpp
index 25b6375ce32..29f401ab7bd 100644
--- a/llvm/lib/Target/X86/X86RegisterInfo.cpp
+++ b/llvm/lib/Target/X86/X86RegisterInfo.cpp
@@ -1149,6 +1149,31 @@ MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI, unsigned OpNu
   return foldMemoryOperand(MI, OpNum, MOs);
 }
 
+MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI,
+                                              SmallVectorImpl<unsigned> &UseOps,
+                                              int FrameIndex) const {
+  // Check switch flag 
+  if (NoFusing) return NULL;
+
+  if (UseOps.size() == 1)
+    return foldMemoryOperand(MI, UseOps[0], FrameIndex);
+  else if (UseOps.size() != 2 || UseOps[0] != 0 && UseOps[1] != 1)
+    return NULL;
+
+  unsigned NewOpc = 0;
+  switch (MI->getOpcode()) {
+  default: return NULL;
+  case X86::TEST8rr:  NewOpc = X86::CMP8ri; break;
+  case X86::TEST16rr: NewOpc = X86::CMP16ri; break;
+  case X86::TEST32rr: NewOpc = X86::CMP32ri; break;
+  case X86::TEST64rr: NewOpc = X86::CMP64ri32; break;
+  }
+  // Change to CMPXXri r, 0 first.
+  MI->setInstrDescriptor(TII.get(NewOpc));
+  MI->getOperand(1).ChangeToImmediate(0);
+  return foldMemoryOperand(MI, 0, FrameIndex);
+}
+
 MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI, unsigned OpNum,
                                                  MachineInstr *LoadMI) const {
   // Check switch flag 
@@ -1160,6 +1185,31 @@ MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI, unsigned OpNu
   return foldMemoryOperand(MI, OpNum, MOs);
 }
 
+MachineInstr* X86RegisterInfo::foldMemoryOperand(MachineInstr *MI,
+                                                 SmallVectorImpl<unsigned> &UseOps,
+                                                 MachineInstr *LoadMI) const {
+  // Check switch flag 
+  if (NoFusing) return NULL;
+
+  if (UseOps.size() == 1)
+    return foldMemoryOperand(MI, UseOps[0], LoadMI);
+  else if (UseOps.size() != 2 || UseOps[0] != 0 && UseOps[1] != 1)
+    return NULL;
+  unsigned NewOpc = 0;
+  switch (MI->getOpcode()) {
+  default: return NULL;
+  case X86::TEST8rr:  NewOpc = X86::CMP8ri; break;
+  case X86::TEST16rr: NewOpc = X86::CMP16ri; break;
+  case X86::TEST32rr: NewOpc = X86::CMP32ri; break;
+  case X86::TEST64rr: NewOpc = X86::CMP64ri32; break;
+  }
+  // Change to CMPXXri r, 0 first.
+  MI->setInstrDescriptor(TII.get(NewOpc));
+  MI->getOperand(1).ChangeToImmediate(0);
+  return foldMemoryOperand(MI, 0, LoadMI);
+}
+
+
 unsigned X86RegisterInfo::getOpcodeAfterMemoryFold(unsigned Opc,
                                                    unsigned OpNum) const {
   // Check switch flag 
@@ -1270,7 +1320,30 @@ bool X86RegisterInfo::unfoldMemoryOperand(MachineFunction &MF, MachineInstr *MI,
     MachineOperand &MO = ImpOps[i];
     MIB.addReg(MO.getReg(), MO.isDef(), true, MO.isKill(), MO.isDead());
   }
-  NewMIs.push_back(MIB);
+  // Change CMP32ri r, 0 back to TEST32rr r, r, etc.
+  unsigned NewOpc = 0;
+  switch (DataMI->getOpcode()) {
+  default: break;
+  case X86::CMP64ri32:
+  case X86::CMP32ri:
+  case X86::CMP16ri:
+  case X86::CMP8ri: {
+    MachineOperand &MO0 = DataMI->getOperand(0);
+    MachineOperand &MO1 = DataMI->getOperand(1);
+    if (MO1.getImm() == 0) {
+      switch (DataMI->getOpcode()) {
+      default: break;
+      case X86::CMP64ri32: NewOpc = X86::TEST64rr; break;
+      case X86::CMP32ri:   NewOpc = X86::TEST32rr; break;
+      case X86::CMP16ri:   NewOpc = X86::TEST16rr; break;
+      case X86::CMP8ri:    NewOpc = X86::TEST8rr; break;
+      }
+      DataMI->setInstrDescriptor(TII.get(NewOpc));
+      MO1.ChangeToRegister(MO0.getReg(), false);
+    }
+  }
+  }
+  NewMIs.push_back(DataMI);
 
   // Emit the store instruction.
   if (UnfoldStore) {
diff --git a/llvm/lib/Target/X86/X86RegisterInfo.h b/llvm/lib/Target/X86/X86RegisterInfo.h
index 18e8b907c3a..53f08440a3f 100644
--- a/llvm/lib/Target/X86/X86RegisterInfo.h
+++ b/llvm/lib/Target/X86/X86RegisterInfo.h
@@ -141,6 +141,12 @@ public:
                                   unsigned OpNum,
                                   int FrameIndex) const;
 
+  /// foldMemoryOperand - Same as previous except it tries to fold instruction
+  /// with multiple uses of the same register.
+  MachineInstr* foldMemoryOperand(MachineInstr* MI,
+                                  SmallVectorImpl<unsigned> &UseOps,
+                                  int FrameIndex) const;
+
   /// foldMemoryOperand - Same as the previous version except it allows folding
   /// of any load and store from / to any address, not just from a specific
   /// stack slot.
@@ -148,6 +154,13 @@ public:
                                   unsigned OpNum,
                                   MachineInstr* LoadMI) const;
 
+  /// foldMemoryOperand - Same as the previous version except it allows folding
+  /// of any load and store from / to any address, not just from a specific
+  /// stack slot.
+  MachineInstr* foldMemoryOperand(MachineInstr* MI,
+                                  SmallVectorImpl<unsigned> &UseOps,
+                                  MachineInstr* LoadMI) const;
+
   /// getOpcodeAfterMemoryFold - Returns the opcode of the would be new
   /// instruction after load / store is folded into an instruction of the
   /// specified opcode. It returns zero if the specified unfolding is not