Prevent ARM assembler from losing a right shift by #32 applied to a register

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

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index e6a0df7b0439135660495f65d830d34f2e7b6611..4f07650e8fc321c4641dd72ec7f15e5acd051c77 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -2304,27 +2304,12 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
     Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                         &MemOpChains[0], MemOpChains.size());
 
-  // Build a sequence of copy-to-reg nodes chained together with token chain
-  // and flag operands which copy the outgoing args into registers.
-  SDValue InFlag;
-  // Tail call byval lowering might overwrite argument registers so in case of
-  // tail call optimization the copies to registers are lowered later.
-  if (!isTailCall)
-    for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) {
-      Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first,
-                               RegsToPass[i].second, InFlag);
-      InFlag = Chain.getValue(1);
-    }
-
   if (Subtarget->isPICStyleGOT()) {
     // ELF / PIC requires GOT in the EBX register before function calls via PLT
     // GOT pointer.
     if (!isTailCall) {
-      Chain = DAG.getCopyToReg(Chain, dl, X86::EBX,
-                               DAG.getNode(X86ISD::GlobalBaseReg,
-                                           DebugLoc(), getPointerTy()),
-                               InFlag);
-      InFlag = Chain.getValue(1);
+      RegsToPass.push_back(std::make_pair(unsigned(X86::EBX),
+               DAG.getNode(X86ISD::GlobalBaseReg, DebugLoc(), getPointerTy())));
     } else {
       // If we are tail calling and generating PIC/GOT style code load the
       // address of the callee into ECX. The value in ecx is used as target of
@@ -2362,12 +2347,10 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
     assert((Subtarget->hasSSE1() || !NumXMMRegs)
            && "SSE registers cannot be used when SSE is disabled");
 
-    Chain = DAG.getCopyToReg(Chain, dl, X86::AL,
-                             DAG.getConstant(NumXMMRegs, MVT::i8), InFlag);
-    InFlag = Chain.getValue(1);
+    RegsToPass.push_back(std::make_pair(unsigned(X86::AL),
+                                        DAG.getConstant(NumXMMRegs, MVT::i8)));
   }
 
-
   // For tail calls lower the arguments to the 'real' stack slot.
   if (isTailCall) {
     // Force all the incoming stack arguments to be loaded from the stack
@@ -2381,8 +2364,6 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
     SmallVector<SDValue, 8> MemOpChains2;
     SDValue FIN;
     int FI = 0;
-    // Do not flag preceding copytoreg stuff together with the following stuff.
-    InFlag = SDValue();
     if (getTargetMachine().Options.GuaranteedTailCallOpt) {
       for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
         CCValAssign &VA = ArgLocs[i];
@@ -2422,19 +2403,20 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
       Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                           &MemOpChains2[0], MemOpChains2.size());
 
-    // Copy arguments to their registers.
-    for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) {
-      Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first,
-                               RegsToPass[i].second, InFlag);
-      InFlag = Chain.getValue(1);
-    }
-    InFlag =SDValue();
-
     // Store the return address to the appropriate stack slot.
     Chain = EmitTailCallStoreRetAddr(DAG, MF, Chain, RetAddrFrIdx, Is64Bit,
                                      FPDiff, dl);
   }
 
+  // Build a sequence of copy-to-reg nodes chained together with token chain
+  // and flag operands which copy the outgoing args into registers.
+  SDValue InFlag;
+  for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) {
+    Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first,
+                             RegsToPass[i].second, InFlag);
+    InFlag = Chain.getValue(1);
+  }
+
   if (getTargetMachine().getCodeModel() == CodeModel::Large) {
     assert(Is64Bit && "Large code model is only legal in 64-bit mode.");
     // In the 64-bit large code model, we have to make all calls
@@ -2536,14 +2518,6 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
     Ops.push_back(DAG.getRegister(RegsToPass[i].first,
                                   RegsToPass[i].second.getValueType()));
 
-  // Add an implicit use GOT pointer in EBX.
-  if (!isTailCall && Subtarget->isPICStyleGOT())
-    Ops.push_back(DAG.getRegister(X86::EBX, getPointerTy()));
-
-  // Add an implicit use of AL for non-Windows x86 64-bit vararg functions.
-  if (Is64Bit && isVarArg && !IsWin64)
-    Ops.push_back(DAG.getRegister(X86::AL, MVT::i8));
-
   // Add a register mask operand representing the call-preserved registers.
   const TargetRegisterInfo *TRI = getTargetMachine().getRegisterInfo();
   const uint32_t *Mask = TRI->getCallPreservedMask(CallConv);
@@ -5047,8 +5021,16 @@ X86TargetLowering::LowerVectorBroadcast(SDValue &Op, SelectionDAG &DAG) const {
 
       SDValue Sc = Op.getOperand(0);
       if (Sc.getOpcode() != ISD::SCALAR_TO_VECTOR &&
-          Sc.getOpcode() != ISD::BUILD_VECTOR)
-        return SDValue();
+          Sc.getOpcode() != ISD::BUILD_VECTOR) {
+
+        if (!Subtarget->hasAVX2())
+          return SDValue();
+
+        // Use the register form of the broadcast instruction available on AVX2.
+        if (VT.is256BitVector())
+          Sc = Extract128BitVector(Sc, 0, DAG, dl);
+        return DAG.getNode(X86ISD::VBROADCAST, dl, VT, Sc);
+      }
 
       Ld = Sc.getOperand(0);
       ConstSplatVal = (Ld.getOpcode() == ISD::Constant ||
@@ -12362,8 +12344,9 @@ X86TargetLowering::EmitLoweredSegAlloca(MachineInstr *MI, MachineBasicBlock *BB,
     BuildMI(mallocMBB, DL, TII->get(X86::MOV64rr), X86::RDI)
       .addReg(sizeVReg);
     BuildMI(mallocMBB, DL, TII->get(X86::CALL64pcrel32))
-      .addExternalSymbol("__morestack_allocate_stack_space").addReg(X86::RDI)
+      .addExternalSymbol("__morestack_allocate_stack_space")
       .addRegMask(RegMask)
+      .addReg(X86::RDI, RegState::Implicit)
       .addReg(X86::RAX, RegState::ImplicitDefine);
   } else {
     BuildMI(mallocMBB, DL, TII->get(X86::SUB32ri), physSPReg).addReg(physSPReg)