Merging r259342 (with s/p2align 4/align 16) because r258750 is not in 3.8.

[oota-llvm.git] / lib / Target / X86 / X86ISelLowering.cpp

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 2aa54e25b0e703594904e73a858a06e5cdb0db49..390c4dae9b62f811c1234b6fa6c219ce6183c376 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1335,6 +1335,7 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
 
     setOperationAction(ISD::BR_CC,              MVT::i1,    Expand);
     setOperationAction(ISD::SETCC,              MVT::i1,    Custom);
+    setOperationAction(ISD::SETCCE,             MVT::i1,    Custom);
     setOperationAction(ISD::SELECT_CC,          MVT::i1,    Expand);
     setOperationAction(ISD::XOR,                MVT::i1,    Legal);
     setOperationAction(ISD::OR,                 MVT::i1,    Legal);
@@ -4171,6 +4172,35 @@ static bool hasFPCMov(unsigned X86CC) {
   }
 }
 
+
+bool X86TargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info,
+                                           const CallInst &I,
+                                           unsigned Intrinsic) const {
+
+  const IntrinsicData* IntrData = getIntrinsicWithChain(Intrinsic);
+  if (!IntrData)
+    return false;
+
+  switch (IntrData->Type) {
+  case LOADA:
+  case LOADU: {
+    Info.opc = ISD::INTRINSIC_W_CHAIN;
+    Info.memVT = MVT::getVT(I.getType());
+    Info.ptrVal = I.getArgOperand(0);
+    Info.offset = 0;
+    Info.align = (IntrData->Type == LOADA ? Info.memVT.getSizeInBits()/8 : 1);
+    Info.vol = false;
+    Info.readMem = true;
+    Info.writeMem = false;
+    return true;
+  }
+  default:
+    break;
+  }
+
+  return false;
+}
+
 /// Returns true if the target can instruction select the
 /// specified FP immediate natively. If false, the legalizer will
 /// materialize the FP immediate as a load from a constant pool.
@@ -14946,8 +14976,11 @@ SDValue X86TargetLowering::LowerSETCCE(SDValue Op, SelectionDAG &DAG) const {
   assert(Carry.getOpcode() != ISD::CARRY_FALSE);
   SDVTList VTs = DAG.getVTList(LHS.getValueType(), MVT::i32);
   SDValue Cmp = DAG.getNode(X86ISD::SBB, DL, VTs, LHS, RHS, Carry);
-  return DAG.getNode(X86ISD::SETCC, DL, Op.getValueType(),
-                     DAG.getConstant(CC, DL, MVT::i8), Cmp.getValue(1));
+  SDValue SetCC = DAG.getNode(X86ISD::SETCC, DL, MVT::i8,
+                              DAG.getConstant(CC, DL, MVT::i8), Cmp.getValue(1));
+  if (Op.getSimpleValueType() == MVT::i1)
+      return DAG.getNode(ISD::TRUNCATE, DL, MVT::i1, SetCC);
+  return SetCC;
 }
 
 // isX86LogicalCmp - Return true if opcode is a X86 logical comparison.
@@ -17429,7 +17462,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget *Subtarget,
       // We need a frame pointer because this will get lowered to a PUSH/POP
       // sequence.
       MachineFrameInfo *MFI = DAG.getMachineFunction().getFrameInfo();
-      MFI->setHasOpaqueSPAdjustment(true);
+      MFI->setHasCopyImplyingStackAdjustment(true);
       // Don't do anything here, we will expand these intrinsics out later
       // during ExpandISelPseudos in EmitInstrWithCustomInserter.
       return SDValue();
@@ -17538,7 +17571,6 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget *Subtarget,
     return DAG.getMergeValues(Results, dl);
   }
   case COMPRESS_TO_MEM: {
-    SDLoc dl(Op);
     SDValue Mask = Op.getOperand(4);
     SDValue DataToCompress = Op.getOperand(3);
     SDValue Addr = Op.getOperand(2);
@@ -17564,7 +17596,6 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget *Subtarget,
   case TRUNCATE_TO_MEM_VI32:
     return LowerINTRINSIC_TRUNCATE_TO_MEM(Op, DAG, MVT::i32);
   case EXPAND_FROM_MEM: {
-    SDLoc dl(Op);
     SDValue Mask = Op.getOperand(4);
     SDValue PassThru = Op.getOperand(3);
     SDValue Addr = Op.getOperand(2);
@@ -17584,6 +17615,25 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget *Subtarget,
                            Mask, PassThru, Subtarget, DAG), Chain};
     return DAG.getMergeValues(Results, dl);
   }
+  case LOADU:
+  case LOADA: {
+    SDValue Mask = Op.getOperand(4);
+    SDValue PassThru = Op.getOperand(3);
+    SDValue Addr = Op.getOperand(2);
+    SDValue Chain = Op.getOperand(0);
+    MVT VT = Op.getSimpleValueType();
+
+    MemIntrinsicSDNode *MemIntr = dyn_cast<MemIntrinsicSDNode>(Op);
+    assert(MemIntr && "Expected MemIntrinsicSDNode!");
+
+    if (isAllOnesConstant(Mask)) // return just a load
+      return DAG.getLoad(VT, dl, Chain, Addr, MemIntr->getMemOperand());
+
+    MVT MaskVT = MVT::getVectorVT(MVT::i1, VT.getVectorNumElements());
+    SDValue VMask = getMaskNode(Mask, MaskVT, Subtarget, DAG, dl);
+    return DAG.getMaskedLoad(VT, dl, Chain, Addr, VMask, PassThru, VT,
+                             MemIntr->getMemOperand(), ISD::NON_EXTLOAD);
+  }
   }
 }
 
@@ -20749,6 +20799,8 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case X86ISD::VSHLI:              return "X86ISD::VSHLI";
   case X86ISD::VSRLI:              return "X86ISD::VSRLI";
   case X86ISD::VSRAI:              return "X86ISD::VSRAI";
+  case X86ISD::VROTLI:             return "X86ISD::VROTLI";
+  case X86ISD::VROTRI:             return "X86ISD::VROTRI";
   case X86ISD::CMPP:               return "X86ISD::CMPP";
   case X86ISD::PCMPEQ:             return "X86ISD::PCMPEQ";
   case X86ISD::PCMPGT:             return "X86ISD::PCMPGT";
@@ -21832,7 +21884,8 @@ X86TargetLowering::EmitLoweredSelect(MachineInstr *MI,
   if (LastCMOV == MI &&
       NextMIIt != BB->end() && NextMIIt->getOpcode() == MI->getOpcode() &&
       NextMIIt->getOperand(2).getReg() == MI->getOperand(2).getReg() &&
-      NextMIIt->getOperand(1).getReg() == MI->getOperand(0).getReg()) {
+      NextMIIt->getOperand(1).getReg() == MI->getOperand(0).getReg() &&
+      NextMIIt->getOperand(1).isKill()) {
     CascadedCMOV = &*NextMIIt;
   }
 
@@ -28860,6 +28913,7 @@ void X86TargetLowering::insertCopiesSplitCSR(
 
   const TargetInstrInfo *TII = Subtarget->getInstrInfo();
   MachineRegisterInfo *MRI = &Entry->getParent()->getRegInfo();
+  MachineBasicBlock::iterator MBBI = Entry->begin();
   for (const MCPhysReg *I = IStart; *I; ++I) {
     const TargetRegisterClass *RC = nullptr;
     if (X86::GR64RegClass.contains(*I))
@@ -28877,13 +28931,13 @@ void X86TargetLowering::insertCopiesSplitCSR(
                Attribute::NoUnwind) &&
            "Function should be nounwind in insertCopiesSplitCSR!");
     Entry->addLiveIn(*I);
-    BuildMI(*Entry, Entry->begin(), DebugLoc(), TII->get(TargetOpcode::COPY),
-            NewVR)
+    BuildMI(*Entry, MBBI, DebugLoc(), TII->get(TargetOpcode::COPY), NewVR)
         .addReg(*I);
 
+    // Insert the copy-back instructions right before the terminator.
     for (auto *Exit : Exits)
-      BuildMI(*Exit, Exit->begin(), DebugLoc(), TII->get(TargetOpcode::COPY),
-              *I)
+      BuildMI(*Exit, Exit->getFirstTerminator(), DebugLoc(),
+              TII->get(TargetOpcode::COPY), *I)
           .addReg(NewVR);
   }
 }