diff --git a/llvm/lib/Target/ARM/ARMFrameLowering.cpp b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
index 196f58dc5eee..79e5de14ddd8 100644
--- a/llvm/lib/Target/ARM/ARMFrameLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
@@ -380,15 +380,10 @@ void ARMFrameLowering::emitEpilogue(MachineFunction &MF,
     if (NumBytes != 0)
       emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);
   } else {
-    MachineBasicBlock::iterator FirstPop = MBBI;
-
     // Unwind MBBI to point to first LDR / VLDRD.
     const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
     if (MBBI != MBB.begin()) {
       do {
-        if (isPopOpcode(MBBI->getOpcode()))
-          FirstPop = MBBI;
-
         --MBBI;
       } while (MBBI != MBB.begin() && isCSRestore(MBBI, TII, CSRegs));
       if (!isCSRestore(MBBI, TII, CSRegs))
@@ -435,7 +430,7 @@ void ARMFrameLowering::emitEpilogue(MachineFunction &MF,
             .addReg(FramePtr));
       }
     } else if (NumBytes &&
-               !tryFoldSPUpdateIntoPushPop(STI, MF, FirstPop, NumBytes))
+               !tryFoldSPUpdateIntoPushPop(STI, MF, MBBI, NumBytes))
         emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);
 
     // Increment past our save areas.
diff --git a/llvm/test/CodeGen/ARM/fold-stack-adjust.ll b/llvm/test/CodeGen/ARM/fold-stack-adjust.ll
index 8bda7683f902..8c60bca1db05 100644
--- a/llvm/test/CodeGen/ARM/fold-stack-adjust.ll
+++ b/llvm/test/CodeGen/ARM/fold-stack-adjust.ll
@@ -124,3 +124,32 @@ define arm_aapcs_vfpcc double @check_vfp_no_return_clobber() minsize {
 
   ret double 1.0
 }
+
+@dbl = global double 0.0
+
+; PR18136: there was a bug determining where the first eligible pop in a
+; basic-block was when the entire block was epilogue code.
+define void @test_fold_point(i1 %tst) minsize {
+; CHECK-LABEL: test_fold_point:
+
+  ; Important to check for beginning of basic block, because if it gets
+  ; if-converted the test is probably no longer checking what it should.
+; CHECK: {{LBB[0-9]+_2}}:
+; CHECK-NEXT: vpop {d7, d8}
+; CHECK-NEXT: pop {r4, pc}
+  ; We want some memory so there's a stack adjustment to fold...
+  %var = alloca i8, i32 8
+
+  ; We want a long-lived floating register so that a callee-saved dN is used and
+  ; there's both a vpop and a pop.
+  %live_val = load double* @dbl
+  br i1 %tst, label %true, label %end
+true:
+  call void @bar(i8* %var)
+  store double %live_val, double* @dbl
+  br label %end
+end:
+  ; We want the epilogue to be the only thing in a basic block so that we hit
+  ; the correct edge-case (first inst in block is correct one to adjust).
+  ret void
+}
\ No newline at end of file