Coalescer should not delete copy instructions whose defs are partially dead. e.g.

%RDI<def,dead> = MOV64rr %RAX<kill>, %EDI<imp-def>

llvm-svn: 100804

llvm/include/llvm/CodeGen/MachineInstr.h

@@ -357,6 +357,10 @@ public:
   /// return 0.
   unsigned isConstantValuePHI() const;
 
+  /// allDefsAreDead - Return true if all the defs of this instruction are dead.
+  ///
+  bool allDefsAreDead() const;
+
   //
   // Debugging support
   //

llvm/lib/CodeGen/MachineInstr.cpp

@@ -1125,6 +1125,19 @@ unsigned MachineInstr::isConstantValuePHI() const {
   return Reg;
 }
 
+/// allDefsAreDead - Return true if all the defs of this instruction are dead.
+///
+bool MachineInstr::allDefsAreDead() const {
+  for (unsigned i = 0, e = getNumOperands(); i < e; ++i) {
+    const MachineOperand &MO = getOperand(i);
+    if (!MO.isReg() || MO.isUse())
+      continue;
+    if (!MO.isDead())
+      return false;
+  }
+  return true;
+}
+
 void MachineInstr::dump() const {
   dbgs() << "  " << *this;
 }

llvm/lib/CodeGen/SimpleRegisterCoalescing.cpp

@@ -2744,7 +2744,7 @@ bool SimpleRegisterCoalescing::runOnMachineFunction(MachineFunction &fn) {
             // delete them later.
             DoDelete = false;
         }
-        if (MI->registerDefIsDead(DstReg)) {
+        if (MI->allDefsAreDead()) {
           LiveInterval &li = li_->getInterval(DstReg);
           if (!ShortenDeadCopySrcLiveRange(li, MI))
             ShortenDeadCopyLiveRange(li, MI);

llvm/test/CodeGen/X86/2010-04-08-CoalescerBug.ll

@@ -0,0 +1,26 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s
+; rdar://7842028
+
+; Do not delete partially dead copy instructions.
+; %RDI<def,dead> = MOV64rr %RAX<kill>, %EDI<imp-def>
+; REP_MOVSD %ECX<imp-def,dead>, %EDI<imp-def,dead>, %ESI<imp-def,dead>, %ECX<imp-use,kill>, %EDI<imp-use,kill>, %ESI<imp-use,kill>
+
+
+%struct.F = type { %struct.FC*, i32, i32, i8, i32, i32, i32 }
+%struct.FC = type { [10 x i8], [32 x i32], %struct.FC*, i32 }
+
+define void @t(%struct.F* %this) nounwind {
+entry:
+; CHECK: t:
+; CHECK: addq $12, %rsi
+  %BitValueArray = alloca [32 x i32], align 4
+  %tmp2 = getelementptr inbounds %struct.F* %this, i64 0, i32 0
+  %tmp3 = load %struct.FC** %tmp2, align 8
+  %tmp4 = getelementptr inbounds %struct.FC* %tmp3, i64 0, i32 1, i64 0
+  %tmp5 = bitcast [32 x i32]* %BitValueArray to i8*
+  %tmp6 = bitcast i32* %tmp4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %tmp5, i8* %tmp6, i64 128, i32 4, i1 false)
+  unreachable
+}
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind