diff --git a/include/llvm/CodeGen/TargetLowering.h b/include/llvm/CodeGen/TargetLowering.h
index fa71024..70fb2f5 100644
--- a/include/llvm/CodeGen/TargetLowering.h
+++ b/include/llvm/CodeGen/TargetLowering.h
@@ -3663,7 +3663,7 @@
                                             SelectionDAG &DAG) const;
 
   // Lower custom output constraints. If invalid, return SDValue().
-  virtual SDValue LowerAsmOutputForConstraint(SDValue &Chain, SDValue *Flag,
+  virtual SDValue LowerAsmOutputForConstraint(SDValue &Chain, SDValue &Flag,
                                               SDLoc DL,
                                               const AsmOperandInfo &OpInfo,
                                               SelectionDAG &DAG) const;
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index c633838..75f7cc0 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -8174,7 +8174,7 @@
             DAG, FuncInfo, getCurSDLoc(), Chain, &Flag, CS.getInstruction());
         break;
       case TargetLowering::C_Other:
-        Val = TLI.LowerAsmOutputForConstraint(Chain, &Flag, getCurSDLoc(),
+        Val = TLI.LowerAsmOutputForConstraint(Chain, Flag, getCurSDLoc(),
                                               OpInfo, DAG);
         break;
       case TargetLowering::C_Memory:
@@ -8185,7 +8185,6 @@
 
       // Indirect output manifest as stores. Record output chains.
       if (OpInfo.isIndirect) {
-
         const Value *Ptr = OpInfo.CallOperandVal;
         assert(Ptr && "Expected value CallOperandVal for indirect asm operand");
         SDValue Store = DAG.getStore(Chain, getCurSDLoc(), Val, getValue(Ptr),
diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index ad2e5e8..143845f 100644
--- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -3280,7 +3280,7 @@
 }
 
 SDValue TargetLowering::LowerAsmOutputForConstraint(
-    SDValue &Chain, SDValue *Flag, SDLoc DL, const AsmOperandInfo &OpInfo,
+    SDValue &Chain, SDValue &Flag, SDLoc DL, const AsmOperandInfo &OpInfo,
     SelectionDAG &DAG) const {
   return SDValue();
 }
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 47c608d..54e8251 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -42733,7 +42733,7 @@
 
 // Lower @cc targets via setcc.
 SDValue X86TargetLowering::LowerAsmOutputForConstraint(
-    SDValue &Chain, SDValue *Flag, SDLoc DL, const AsmOperandInfo &OpInfo,
+    SDValue &Chain, SDValue &Flag, SDLoc DL, const AsmOperandInfo &OpInfo,
     SelectionDAG &DAG) const {
   X86::CondCode Cond = parseConstraintCode(OpInfo.ConstraintCode);
   if (Cond == X86::COND_INVALID)
@@ -42744,14 +42744,13 @@
     report_fatal_error("Flag output operand is of invalid type");
 
   // Get EFLAGS register. Only update chain when copyfrom is glued.
-  SDValue EFlags;
-  if (Flag) {
-    EFlags = DAG.getCopyFromReg(Chain, DL, X86::EFLAGS, MVT::i32, *Flag);
-    Chain = EFlags.getValue(1);
+  if (Flag.getNode()) {
+    Flag = DAG.getCopyFromReg(Chain, DL, X86::EFLAGS, MVT::i32, Flag);
+    Chain = Flag.getValue(1);
   } else
-    EFlags = DAG.getCopyFromReg(Chain, DL, X86::EFLAGS, MVT::i32);
+    Flag = DAG.getCopyFromReg(Chain, DL, X86::EFLAGS, MVT::i32);
   // Extract CC code.
-  SDValue CC = getSETCC(Cond, EFlags, DL, DAG);
+  SDValue CC = getSETCC(Cond, Flag, DL, DAG);
   // Extend to 32-bits
   SDValue Result = DAG.getNode(ISD::ZERO_EXTEND, DL, OpInfo.ConstraintVT, CC);
 
diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index 45d7308..a328b95 100644
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -922,7 +922,7 @@
     }
 
     /// Handle Lowering flag assembly outputs.
-    SDValue LowerAsmOutputForConstraint(SDValue &Chain, SDValue *Flag, SDLoc DL,
+    SDValue LowerAsmOutputForConstraint(SDValue &Chain, SDValue &Flag, SDLoc DL,
                                         const AsmOperandInfo &Constraint,
                                         SelectionDAG &DAG) const override;
 
diff --git a/test/CodeGen/X86/pr40737.ll b/test/CodeGen/X86/pr40737.ll
new file mode 100644
index 0000000..2643b78
--- /dev/null
+++ b/test/CodeGen/X86/pr40737.ll
@@ -0,0 +1,19 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu %s -o - | FileCheck %s
+
+define i8 @_BitScanForward(i32* nocapture %Index, i32 %Mask) {
+; CHECK-LABEL: _BitScanForward:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    #APP
+; CHECK-NEXT:    bsfl %esi, %ecx
+; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:    setne %al
+; CHECK-NEXT:    movl %ecx, (%rdi)
+; CHECK-NEXT:    retq
+entry:
+  %0 = tail call { i8, i32 } asm "bsf$(l $2,$1 $| $1,$2$)", "={@ccnz},=r,r,~{dirflag},~{fpsr},~{flags}"(i32 %Mask)
+  %asmresult = extractvalue { i8, i32 } %0, 0
+  %asmresult1 = extractvalue { i8, i32 } %0, 1
+  store i32 %asmresult1, i32* %Index, align 4
+  ret i8 %asmresult
+}