diff --git a/bolt/include/bolt/Core/MCPlusBuilder.h b/bolt/include/bolt/Core/MCPlusBuilder.h
--- a/bolt/include/bolt/Core/MCPlusBuilder.h
+++ b/bolt/include/bolt/Core/MCPlusBuilder.h
@@ -110,6 +110,13 @@
     return AnnotationInst;
   }
 
+  void removeAnnotationInst(MCInst &Inst) const {
+    assert(getAnnotationInst(Inst) && "Expected annotation instruction.");
+    Inst.erase(std::prev(Inst.end()));
+    assert(!getAnnotationInst(Inst) &&
+           "More than one annotation instruction detected.");
+  }
+
   void setAnnotationOpValue(MCInst &Inst, unsigned Index, int64_t Value,
                             AllocatorIdTy AllocatorId = 0) {
     MCInst *AnnotationInst = getAnnotationInst(Inst);
@@ -163,6 +170,18 @@
   /// MCPlusBuilder classes must use convert/lower/create* interfaces instead.
   void setTailCall(MCInst &Inst);
 
+  /// Transfer annotations from \p SrcInst to \p DstInst.
+  void moveAnnotations(MCInst &&SrcInst, MCInst &DstInst) const {
+    assert(!getAnnotationInst(DstInst) &&
+           "Destination instruction should not have annotations.");
+    const MCInst *AnnotationInst = getAnnotationInst(SrcInst);
+    if (!AnnotationInst)
+      return;
+
+    DstInst.addOperand(MCOperand::createInst(AnnotationInst));
+    removeAnnotationInst(SrcInst);
+  }
+
 public:
   class InstructionIterator {
   public:
@@ -1864,9 +1883,8 @@
   void stripAnnotations(MCInst &Inst, bool KeepTC = false);
 
   virtual InstructionListType
-  createInstrumentedIndirectCall(const MCInst &CallInst, bool TailCall,
-                                 MCSymbol *HandlerFuncAddr, int CallSiteID,
-                                 MCContext *Ctx) {
+  createInstrumentedIndirectCall(MCInst &&CallInst, MCSymbol *HandlerFuncAddr,
+                                 int CallSiteID, MCContext *Ctx) {
     llvm_unreachable("not implemented");
     return InstructionListType();
   }
diff --git a/bolt/lib/Core/MCPlusBuilder.cpp b/bolt/lib/Core/MCPlusBuilder.cpp
--- a/bolt/lib/Core/MCPlusBuilder.cpp
+++ b/bolt/lib/Core/MCPlusBuilder.cpp
@@ -298,7 +298,8 @@
   // Preserve TailCall annotation.
   auto IsTC = hasAnnotation(Inst, MCAnnotation::kTailCall);
 
-  Inst.erase(std::prev(Inst.end()));
+  removeAnnotationInst(Inst);
+
   if (KeepTC && IsTC)
     setTailCall(Inst);
 }
diff --git a/bolt/lib/Passes/Instrumentation.cpp b/bolt/lib/Passes/Instrumentation.cpp
--- a/bolt/lib/Passes/Instrumentation.cpp
+++ b/bolt/lib/Passes/Instrumentation.cpp
@@ -215,7 +215,7 @@
   BinaryContext &BC = FromFunction.getBinaryContext();
   bool IsTailCall = BC.MIB->isTailCall(*Iter);
   InstructionListType CounterInstrs = BC.MIB->createInstrumentedIndirectCall(
-      *Iter, IsTailCall,
+      std::move(*Iter),
       IsTailCall ? IndTailCallHandlerExitBBFunction->getSymbol()
                  : IndCallHandlerExitBBFunction->getSymbol(),
       IndCallSiteID, &*BC.Ctx);
diff --git a/bolt/lib/Target/X86/X86MCPlusBuilder.cpp b/bolt/lib/Target/X86/X86MCPlusBuilder.cpp
--- a/bolt/lib/Target/X86/X86MCPlusBuilder.cpp
+++ b/bolt/lib/Target/X86/X86MCPlusBuilder.cpp
@@ -3085,8 +3085,7 @@
     Inst.addOperand(MCOperand::createReg(X86::NoRegister)); // AddrSegmentReg
   }
 
-  InstructionListType createInstrumentedIndirectCall(const MCInst &CallInst,
-                                                     bool TailCall,
+  InstructionListType createInstrumentedIndirectCall(MCInst &&CallInst,
                                                      MCSymbol *HandlerFuncAddr,
                                                      int CallSiteID,
                                                      MCContext *Ctx) override {
@@ -3137,14 +3136,13 @@
     createLoadImmediate(Insts.back(), TempReg, CallSiteID);
     Insts.emplace_back();
     createPushRegister(Insts.back(), TempReg, 8);
-    Insts.emplace_back();
-    createDirectCall(Insts.back(), HandlerFuncAddr, Ctx,
-                     /*TailCall=*/TailCall);
-    // Carry over metadata
-    for (int I = MCPlus::getNumPrimeOperands(CallInst),
-             E = CallInst.getNumOperands();
-         I != E; ++I)
-      Insts.back().addOperand(CallInst.getOperand(I));
+
+    MCInst &NewCallInst = Insts.emplace_back();
+    createDirectCall(NewCallInst, HandlerFuncAddr, Ctx, isTailCall(CallInst));
+
+    // Carry over metadata including tail call marker if present.
+    stripAnnotations(NewCallInst);
+    moveAnnotations(std::move(CallInst), NewCallInst);
 
     return Insts;
   }